Visual Acuity Test for LLMs

Test Image (4K)

Click image to expand/collapse

Leaderboard (4K)

Rank	Model	Provider	Input Tokens	Output Tokens	Threshold Row	Threshold Size	Mean Accuracy	Attempted Rows
1T	GPT-5.4 (original detail)	OpenAI	12,045	198	12	6px	100.0%	12/12
2T	GPT-5.5 (none)	OpenAI	12,045	198	12	6px	100.0%	12/12
3T	GPT-5.6 (none)	OpenAI	19,705	198	12	6px	100.0%	12/12
4T	GPT-5.6-LUNA (none)	OpenAI	19,705	198	12	6px	100.0%	12/12
5T	GPT-5.6-TERRA (none)	OpenAI	19,705	185	12	6px	100.0%	12/12
6	GPT-5.6 (none) (high detail)	OpenAI	3,045	198	12	6px	96.7%	12/12
7	Claude Opus 4 7 (no thinking) (effort: low)	Anthropic	4,822	312	12	6px	91.7%	12/12
8T	Gemini 3 Flash Preview (thinking: minimal) (high res)	Gemini	1,129	242	11	8px	95.0%	12/12
9T	Gemini 3 Pro Preview (high res)	Gemini	1,129	242	11	8px	95.0%	12/12
10	Claude Opus 4 8 (no thinking) (effort: low)	Anthropic	4,816	313	10	10px	90.0%	12/12
11	GPT-5-MINI (minimal)	OpenAI	1,870	199	10	10px	89.1%	11/12
12	GPT-4.1-NANO (high detail)	OpenAI	3,788	160	10	10px	88.0%	10/12
13T	GPT-4.1-MINI	OpenAI	2,511	183	10	10px	86.7%	12/12
14T	GPT-4.1-MINI (low detail)	OpenAI	2,511	195	10	10px	86.7%	12/12
15	GPT-4.1-MINI (high detail)	OpenAI	2,511	200	10	10px	85.0%	12/12
16T	Claude Fable 5 (effort: low)	Anthropic	4,816	355	10	10px	83.3%	12/12
17T	Claude Sonnet 5 (no thinking) (effort: low)	Anthropic	4,816	332	10	10px	83.3%	12/12
18	Grok 4.5 (low)	xAI	2,649	220	10	10px	81.7%	12/12
19	Gemini 3 Pro Preview (medium res)	Gemini	569	198	9	12px	97.8%	9/12
20	GPT-4.1-NANO (low detail)	OpenAI	3,788	165	9	12px	90.0%	10/12
21	Gemini 3 Flash Preview (thinking: minimal) (medium res)	Gemini	569	212	8	15px	73.3%	12/12
22	Gemini 3 Pro Preview (low res)	Gemini	296	185	7	19px	87.5%	8/12
23	Claude Haiku 4 5 20251001	Anthropic	1,571	226	7	19px	74.0%	10/12
24	Gemini 2.5 Pro	Gemini	298	214	7	19px	61.8%	11/12
25	Gemini 3 Flash Preview (thinking: minimal) (low res)	Gemini	296	265	7	19px	61.7%	12/12
26T	Claude Sonnet 4 5 20250929	Anthropic	1,571	292	7	19px	60.0%	12/12
27T	Claude Sonnet 4 6 (no thinking) (effort: low)	Anthropic	1,571	314	7	19px	60.0%	12/12
28	GPT-4.1	OpenAI	810	272	7	19px	58.3%	12/12
29T	Claude Opus 4 5 20251101	Anthropic	1,571	289	7	19px	56.7%	12/12
30T	Claude Opus 4 6 (no thinking) (effort: low)	Anthropic	1,571	221	7	19px	56.7%	12/12
31T	GPT-5 (minimal)	OpenAI	674	225	7	19px	56.7%	12/12
32	GPT-4O (high detail)	OpenAI	810	167	7	19px	55.0%	12/12
33	Gemini 2.0 Flash (high res)	Gemini	3,393	122	6	24px	100.0%	6/12
34	Claude Sonnet 4 20250514	Anthropic	1,571	206	6	24px	94.3%	7/12
35	Gemini 2.0 Flash (medium res)	Gemini	297	180	6	24px	71.1%	9/12
36	Claude Opus 4 1 20250805	Anthropic	1,571	296	6	24px	64.0%	10/12
37	Gemini 2.5 Flash (no thinking) (medium res)	Gemini	298	252	6	24px	56.7%	12/12
38	Gemini 2.5 Pro (low res)	Gemini	106	136	6	24px	56.4%	11/12
39T	Gemini 2.0 Flash (low res)	Gemini	105	164	6	24px	55.0%	12/12
40T	Gemini 2.5 Flash (no thinking)	Gemini	298	347	6	24px	55.0%	12/12
41T	Gemini 2.5 Pro (medium res)	Gemini	298	239	6	24px	55.0%	12/12
42	GPT-4O-MINI (high detail)	OpenAI	25,546	302	6	24px	51.7%	12/12
43	GPT-4O-MINI (low detail)	OpenAI	2,878	145	4	38px	42.0%	10/12
44	Grok 4 0709	xAI	978	60	3	48px	85.0%	4/12
45T	GPT-5.5 (none) (low detail)	OpenAI	352	248	2	60px	16.7%	12/12
46T	GPT-5.6 (none) (low detail)	OpenAI	352	246	2	60px	16.7%	12/12
47	GPT-4O (low detail)	OpenAI	130	22	None	N/A	0.0%	0/12

Click a row to see detailed results
Use ← → arrow keys to navigate between models when a card is open

1T GPT-5.4 (original detail)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: original, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

2T GPT-5.5 (none)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

3T GPT-5.6 (none)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

4T GPT-5.6-LUNA (none)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

5T GPT-5.6-TERRA (none)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[["S", "D", "C", "D", "Z"], ["C", "D", "D", "O", "K"], ["C", "C", "C", "O", "Z"], ["S", "Z", "S", "H", "D"], ["H", "D", "C", "V", "S"], ["K", "D", "V", "S", "D"], ["Z", "D", "V", "Z", "C"], ["N", "V", "D", "N", "S"], ["O", "N", "K", "K", "S"], ["V", "C", "V", "D", "C"], ["O", "H", "S", "R", "H"], ["C", "D", "Z", "S", "H"]]

6 GPT-5.6 (none) (high detail)

OpenAI

Threshold Row

96.7%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	O	H	S	K	H	O	H	S	R	H	4/5 (80%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "O", "C"],
  ["O", "H", "S", "K", "H"],
  ["C", "D", "Z", "S", "H"]
]

7 Claude Opus 4 7 (no thinking) (effort: low)

Anthropic

Threshold Row

91.7%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: max_tokens: 2000, thinking_mode: disabled, effort: low

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	O	N	S	K	N	O	H	S	R	H	2/5 (40%)
12	6px	C	D	Z	S	N	C	D	Z	S	H	4/5 (80%)

Full Response:

Here are the letters row by row:

```python
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"],
    ["Z", "D", "V", "Z", "C"],
    ["N", "V", "D", "N", "S"],
    ["O", "N", "K", "K", "S"],
    ["V", "C", "V", "O", "C"],
    ["O", "N", "S", "K", "N"],  # not sure
    ["C", "D", "Z", "S", "N"]   # not sure
]
```

The last two rows are very small and blurry, so I'm **not sure** about those letters. The rows above that are legible with reasonable confidence.

8T Gemini 3 Flash Preview (thinking: minimal) (high res)

Gemini

Threshold Row

95.0%

Mean Accuracy

12/12

Attempted Rows

8px

Smallest Letters

Configuration: media_resolution: HIGH, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	K	H	O	H	S	R	H	4/5 (80%)
12	6px	C	O	Z	K	H	C	D	Z	S	H	3/5 (60%)

Full Response:

Based on the image provided, here are the letters row by row. Please note that as the letters get smaller in the lower rows, they become increasingly difficult to distinguish clearly.

*   **Row 1:** S, D, C, D, Z
*   **Row 2:** C, D, D, O, K
*   **Row 3:** C, C, C, O, Z
*   **Row 4:** S, Z, S, H, D
*   **Row 5:** H, D, C, V, S
*   **Row 6:** K, D, V, S, D
*   **Row 7:** Z, D, V, Z, C
*   **Row 8:** N, V, D, N, S
*   **Row 9:** O, N, K, K, S
*   **Row 10:** V, C, V, D, C
*   **Row 11:** O, H, S, K, H
*   **Row 12:** C, O, Z, K, H

9T Gemini 3 Pro Preview (high res)

Gemini

Threshold Row

95.0%

Mean Accuracy

12/12

Attempted Rows

8px

Smallest Letters

Configuration: media_resolution: HIGH, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	K	H	O	H	S	R	H	4/5 (80%)
12	6px	C	O	Z	V	H	C	D	Z	S	H	3/5 (60%)

Full Response:

Based on the image provided, here are the letters row by row:

1.  ['S', 'D', 'C', 'D', 'Z']
2.  ['C', 'D', 'D', 'O', 'K']
3.  ['C', 'C', 'C', 'O', 'Z']
4.  ['S', 'Z', 'S', 'H', 'D']
5.  ['H', 'D', 'C', 'V', 'S']
6.  ['K', 'D', 'V', 'S', 'D']
7.  ['Z', 'D', 'V', 'Z', 'C']
8.  ['N', 'V', 'D', 'N', 'S']
9.  ['O', 'N', 'K', 'K', 'S']
10. ['V', 'C', 'V', 'D', 'C']
11. ['O', 'H', 'S', 'K', 'H']
12. ['C', 'O', 'Z', 'V', 'H']

10 Claude Opus 4 8 (no thinking) (effort: low)

Anthropic

Threshold Row

90.0%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: max_tokens: 2000, thinking_mode: disabled, effort: low

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	O	N	S	N	N	O	H	S	R	H	2/5 (40%)
12	6px	C	O	Z	S	N	C	D	Z	S	H	3/5 (60%)

Full Response:

Reading the letters row by row:

```
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"],
    ["Z", "D", "V", "Z", "C"],
    ["N", "V", "D", "N", "S"],
    ["O", "N", "K", "K", "S"],
    ["V", "C", "V", "O", "C"],
    ["O", "N", "S", "N", "N"],
    ["C", "O", "Z", "S", "N"]
]
```

Note: The last few rows become increasingly faint and small, so several entries there are "not sure" — particularly rows 11 and 12, where the letters are very difficult to read with confidence.

11 GPT-5-MINI (minimal)

OpenAI

Threshold Row

89.1%

Mean Accuracy

11/12

Attempted Rows

10px

Smallest Letters

Configuration: reasoning_effort: minimal, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[
["S", "D", "C", "D", "Z"],
["C", "D", "D", "O", "K"],
["C", "C", "C", "O", "Z"],
["S", "Z", "S", "H", "D"],
["H", "D", "C", "V", "S"],
["K", "D", "V", "S", "D"],
["Z", "D", "V", "Z", "C"],
["N", "V", "D", "N", "S"],
["O", "N", "K", "K", "S"],
["V", "C", "V", "O", "C"],
[".", "?", "?", ".", "."]
]

Note: The smallest/faint letters near the bottom are hard to read; I marked uncertain ones as "?" and a "." where I couldn't discern a letter.

12 GPT-4.1-NANO (high detail)

OpenAI

Threshold Row

88.0%

Mean Accuracy

10/12

Attempted Rows

10px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	D	K	C	D	D	O	K	4/5 (80%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	2	D	V	2	C	Z	D	V	Z	C	3/5 (60%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	S	?	O	N	K	K	S	3/5 (60%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["c", "D", "D", "D", "K"],
  ["c", "c", "c", "o", "Z"],
  ["s", "z", "s", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["2", "D", "V", "2", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "S"],
  ["V", "C", "V", "O", "C"]
]

13T GPT-4.1-MINI

OpenAI

Threshold Row

86.7%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	O	?	S	?	?	O	H	S	R	H	2/5 (40%)
12	6px	C	?	?	?	?	C	D	Z	S	H	1/5 (20%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "O", "C"],
  ["O", "-", "S", "-", "-"],
  ["C", "-", "-", "-", "-"]
]

14T GPT-4.1-MINI (low detail)

OpenAI

Threshold Row

86.7%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	O	S	S	S	O	O	H	S	R	H	2/5 (40%)
12	6px	C	O	I	I	T	C	D	Z	S	H	1/5 (20%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "O", "C"],
  ["O", "S", "S", "S", "O"],
  ["C", "O", "I", "I", "T"]
]

15 GPT-4.1-MINI (high detail)

OpenAI

Threshold Row

85.0%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	O	C	V	C	V	D	C	4/5 (80%)
11	8px	?	?	S	?	?	O	H	S	R	H	1/5 (20%)
12	6px	C	E	J	L	?	C	D	Z	S	H	1/5 (20%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "O", "C"],
  ["not sure", "not sure", "S", "not sure", "not sure"],
  ["C", "E", "J", "L", "not sure"]
]

16T Claude Fable 5 (effort: low)

Anthropic

Threshold Row

83.3%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: max_tokens: 2000, effort: low

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's what I can read, row by row (letters get progressively smaller like an eye chart):

1. [S, D, C, D, Z]
2. [C, D, D, O, K]
3. [C, C, C, O, Z]
4. [S, Z, S, H, D]
5. [H, D, C, V, S]
6. [K, D, V, S, D]
7. [Z, D, V, Z, C]
8. [N, V, D, N, S]
9. [O, N, K, K, S]
10. [V, C, V, D, C]
11. [not sure, not sure, not sure, not sure, not sure] — too small to read reliably
12. [not sure, not sure, not sure, not sure, not sure] — too small to read reliably

17T Claude Sonnet 5 (no thinking) (effort: low)

Anthropic

Threshold Row

83.3%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: max_tokens: 2000, thinking_mode: disabled, effort: low

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here is the list of lists representing the letters row by row, from largest/clearest to smallest/least visible:

```python
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"],
    ["Z", "D", "V", "Z", "C"],
    ["N", "V", "D", "N", "S"],
    ["O", "N", "K", "K", "S"],
    ["V", "C", "V", "D", "C"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"]
]
```

Note: The last two rows contain extremely small text that is too tiny to reliably distinguish individual letters.

18 Grok 4.5 (low)

xAI

Threshold Row

81.7%

Mean Accuracy

12/12

Attempted Rows

10px

Smallest Letters

Configuration: reasoning_effort: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	S	C	V	C	V	D	C	4/5 (80%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

```html
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "S", "C"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]
```

19 Gemini 3 Pro Preview (medium res)

Gemini

Threshold Row

97.8%

Mean Accuracy

9/12

Attempted Rows

12px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	H	S	O	N	K	K	S	4/5 (80%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Based on the visual acuity chart provided, here are the letters row by row. The letters decrease in size, and the last few rows appear as dots which are not legible letters, so they are excluded.

[
  ['S', 'D', 'C', 'D', 'Z'],
  ['C', 'D', 'D', 'O', 'K'],
  ['C', 'C', 'C', 'O', 'Z'],
  ['S', 'Z', 'S', 'H', 'D'],
  ['H', 'D', 'C', 'V', 'S'],
  ['K', 'D', 'V', 'S', 'D'],
  ['Z', 'D', 'V', 'Z', 'C'],
  ['N', 'V', 'D', 'N', 'S'],
  ['O', 'N', 'K', 'H', 'S']
]

20 GPT-4.1-NANO (low detail)

OpenAI

Threshold Row

90.0%

Mean Accuracy

10/12

Attempted Rows

12px

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	2	D	V	2	C	Z	D	V	Z	C	3/5 (60%)
8	15px	?	V	D	N	S	N	V	D	N	S	4/5 (80%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	?	C	V	O	C	V	C	V	D	C	3/5 (60%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[
  ["s", "D", "c", "D", "Z"],
  ["c", "D", "D", "O", "K"],
  ["c", "c", "c", "o", "Z"],
  ["s", "z", "s", "H", "D"],
  ["H", "D", "c", "V", "S"],
  ["K", "D", "v", "s", "D"],
  ["2", "D", "v", "2", "c"],
  ["not sure", "v", "D", "N", "s"],
  ["o", "N", "K", "K", "S"],
  ["not sure", "c", "v", "o", "c"]
]

21 Gemini 3 Flash Preview (thinking: minimal) (medium res)

Gemini

Threshold Row

73.3%

Mean Accuracy

12/12

Attempted Rows

15px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	R	K	R	S	O	N	K	K	S	3/5 (60%)
10	10px	Z	H	V	S	H	V	C	V	D	C	1/5 (20%)
11	8px	N	D	O	V	C	O	H	S	R	H	0/5 (0%)
12	6px	S	K	V	D	S	C	D	Z	S	H	0/5 (0%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "R", "K", "R", "S"],
  ["Z", "H", "V", "S", "H"],
  ["N", "D", "O", "V", "C"],
  ["S", "K", "V", "D", "S"]
]
```

22 Gemini 3 Pro Preview (low res)

Gemini

Threshold Row

87.5%

Mean Accuracy

8/12

Attempted Rows

19px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Based on the provided image, which appears to be a visual acuity chart with rows of decreasing size, here are the letters row by row. The letters become illegible after the 7th row.

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]

23 Claude Haiku 4 5 20251001

Anthropic

Threshold Row

74.0%

Mean Accuracy

10/12

Attempted Rows

19px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	O	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	H	V	U	A	S	N	V	D	N	S	2/5 (40%)
9	12px	C	L	K	L	A	O	N	K	K	S	1/5 (20%)
10	10px	L	L	L	L	L	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

# Letters by Row

```
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "O", "V", "Z", "C"],
  ["H", "V", "U", "A", "S"],
  ["C", "L", "K", "L", "A"],
  ["L", "L", "L", "L", "L"]
]
```

Note: Some letters in the lower rows are very faint and small, making them difficult to read with complete certainty. If you need clarification on any specific letters, please let me know.

24 Gemini 2.5 Pro

Gemini

Threshold Row

61.8%

Mean Accuracy

11/12

Attempted Rows

19px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	O	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

```
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "O", "V", "Z", "C"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]
```

25 Gemini 3 Flash Preview (thinking: minimal) (low res)

Gemini

Threshold Row

61.7%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	H	S	O	N	S	N	V	D	N	S	2/5 (40%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Based on the image provided, here are the letters organized by row. Note that as the rows progress, the font size decreases significantly, making the later rows more difficult to read accurately.

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["H", "S", "O", "N", "S"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]

26T Claude Sonnet 4 5 20250929

Anthropic

Threshold Row

60.0%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	A	D	K	D	V	S	D	4/5 (80%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	H	?	D	A	S	N	V	D	N	S	2/5 (40%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Reading the letters row by row from the image:

```
[
  ['S', 'D', 'C', 'D', 'Z'],
  ['C', 'D', 'D', 'O', 'K'],
  ['C', 'C', 'C', 'O', 'Z'],
  ['s', 'Z', 's', 'H', 'D'],
  ['H', 'D', 'C', 'V', 'S'],
  ['K', 'D', 'V', 'a', 'D'],
  ['Z', 'D', 'V', 'Z', 'C'],
  ['H', 'not sure', 'D', 'a', 's'],
  ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
  ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
  ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
  ['not sure', 'not sure', 'not sure', 'not sure', 'not sure']
]
```

27T Claude Sonnet 4 6 (no thinking) (effort: low)

Anthropic

Threshold Row

60.0%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: max_tokens: 2000, thinking_mode: disabled, effort: low

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	O	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	H	V	O	H	S	N	V	D	N	S	2/5 (40%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here is the list of lists, row by row:

```
Row 1:  ["S", "D", "C", "D", "Z"]
Row 2:  ["C", "D", "D", "O", "K"]
Row 3:  ["C", "C", "C", "O", "Z"]
Row 4:  ["S", "Z", "S", "H", "D"]
Row 5:  ["H", "D", "C", "V", "S"]
Row 6:  ["K", "D", "V", "S", "D"]
Row 7:  ["Z", "O", "V", "Z", "C"]
Row 8:  ["H", "V", "O", "H", "S"]
Row 9:  ["not sure", "not sure", "not sure", "not sure", "not sure"]
Row 10: ["not sure", "not sure", "not sure", "not sure", "not sure"]
Row 11: ["not sure", "not sure", "not sure", "not sure", "not sure"]
Row 12: ["not sure", "not sure", "not sure", "not sure", "not sure"]
```

Rows 9–12 contain characters that are extremely small and faint, making them too difficult to read confidently.

28 GPT-4.1

OpenAI

Threshold Row

58.3%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here is the list of lists representing the letters I see in the image, row by row:

```
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"],
    ["Z", "D", "V", "Z", "C"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"]
]
```

The text becomes difficult to distinguish in the lower rows, so I marked them as "not sure". Let me know if you need anything else!

29T Claude Opus 4 5 20251101

Anthropic

Threshold Row

56.7%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	O	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Looking at the image row by row, here are the letters I can identify:

```python
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"],
    ["Z", "O", "V", "Z", "C"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"],
    ["not sure", "not sure", "not sure", "not sure", "not sure"]
]
```

The letters become increasingly faint/small in the lower rows (rows 8-12), making them very difficult or impossible to read with certainty.

30T Claude Opus 4 6 (no thinking) (effort: low)

Anthropic

Threshold Row

56.7%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: max_tokens: 2000, thinking_mode: disabled, effort: low

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	O	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[["S", "D", "C", "D", "Z"],
["C", "D", "D", "O", "K"],
["C", "C", "C", "O", "Z"],
["S", "Z", "S", "H", "D"],
["H", "D", "C", "V", "S"],
["K", "D", "V", "S", "D"],
["Z", "O", "V", "Z", "C"],
["not sure", "not sure", "not sure", "not sure", "not sure"],
["not sure", "not sure", "not sure", "not sure", "not sure"],
["not sure", "not sure", "not sure", "not sure", "not sure"],
["not sure", "not sure", "not sure", "not sure", "not sure"],
["not sure", "not sure", "not sure", "not sure", "not sure"]]

31T GPT-5 (minimal)

OpenAI

Threshold Row

56.7%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: reasoning_effort: minimal, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	S	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "S", "V", "Z", "C"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]

32 GPT-4O (high detail)

OpenAI

Threshold Row

55.0%

Mean Accuracy

12/12

Attempted Rows

19px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	O	V	S	D	K	D	V	S	D	4/5 (80%)
7	19px	Z	O	V	Z	C	Z	D	V	Z	C	4/5 (80%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

```python
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'D', 'O', 'K'],
    ['C', 'C', 'C', 'O', 'Z'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['H', 'D', 'C', 'V', 'S'],
    ['K', 'O', 'V', 'S', 'D'],
    ['Z', 'O', 'V', 'Z', 'C'],
    ['.', '.', '.', '.', '.'],
    ['.', '.', '.', '.', '.'],
    ['.', '.', '.', '.', '.'],
    ['.', '.', '.', '.', '.'],
    ['.', '.', '.', '.', '.'],
    ['.', '.', '.', '.', '.']
]
```

33 Gemini 2.0 Flash (high res)

Gemini

Threshold Row

100.0%

Mean Accuracy

6/12

Attempted Rows

24px

Smallest Letters

Configuration: media_resolution: HIGH, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	?	?	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here are the letters in the image, row by row:

```
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"]
]
```

34 Claude Sonnet 4 20250514

Anthropic

Threshold Row

94.3%

Mean Accuracy

7/12

Attempted Rows

24px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	C	V	Z	V	Z	D	V	Z	C	3/5 (60%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Looking at the image row by row, I can see the following letters:

[['S', 'D', 'C', 'D', 'Z'],
['C', 'D', 'D', 'O', 'K'],
['C', 'C', 'C', 'O', 'Z'],
['S', 'Z', 'S', 'H', 'D'],
['H', 'D', 'C', 'V', 'S'],
['K', 'D', 'V', 'S', 'D'],
['Z', 'C', 'V', 'Z', 'V']]

Note: The letters become increasingly faint in the lower rows, and some of the very faint letters in the bottom rows I'm less certain about, but I've provided my best reading of what appears to be visible.

35 Gemini 2.0 Flash (medium res)

Gemini

Threshold Row

71.1%

Mean Accuracy

9/12

Attempted Rows

24px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	?	?	V	Z	?	Z	D	V	Z	C	2/5 (40%)
8	15px	V	D	N	V	D	N	V	D	N	S	0/5 (0%)
9	12px	N	C	S	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's the OCR output of the image, represented as a list of lists, row by row:

```
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'D', 'O', 'K'],
    ['C', 'C', 'C', 'O', 'Z'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['H', 'D', 'C', 'V', 'S'],
    ['K', 'D', 'V', 'S', 'D'],
    ['not sure', 'not sure', 'V', 'Z', 'not sure'],
    ['V', 'D', 'N', 'V', 'D'],
    ['N', 'C', 'S']
]
```

36 Claude Opus 4 1 20250805

Anthropic

Threshold Row

64.0%

Mean Accuracy

10/12

Attempted Rows

24px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	Y	S	D	K	D	V	S	D	4/5 (80%)
7	19px	Z	D	Y	Z	D	Z	D	V	Z	C	3/5 (60%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Looking at this image row by row, I can see letters arranged in a grid pattern. Here's what I observe:

```python
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'D', 'O', 'K'],
    ['C', 'C', 'C', 'O', 'Z'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['H', 'D', 'C', 'V', 'S'],
    ['K', 'D', 'Y', 'S', 'D'],
    ['Z', 'D', 'Y', 'Z', 'D'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure']
]
```

The first 7 rows contain clearly visible letters, while the last 3 rows appear to have very faint or unclear characters that I cannot confidently identify.

37 Gemini 2.5 Flash (no thinking) (medium res)

Gemini

Threshold Row

56.7%

Mean Accuracy

12/12

Attempted Rows

24px

Smallest Letters

Configuration: thinking_budget: 0, media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	S	Z	V	C	Z	D	V	Z	C	2/5 (40%)
8	15px	O	D	K	V	X	N	V	D	N	S	0/5 (0%)
9	12px	C	V	S	K	O	O	N	K	K	S	1/5 (20%)
10	10px	H	C	D	V	O	V	C	V	D	C	1/5 (20%)
11	8px	S	Z	H	D	S	O	H	S	R	H	0/5 (0%)
12	6px	Z	S	S	Z	D	C	D	Z	S	H	0/5 (0%)

Full Response:

Here are the letters, row by row, as requested:

[['S', 'D', 'C', 'D', 'Z'],
 ['C', 'D', 'D', 'O', 'K'],
 ['C', 'C', 'C', 'O', 'Z'],
 ['S', 'Z', 'S', 'H', 'D'],
 ['H', 'D', 'C', 'V', 'S'],
 ['K', 'D', 'V', 'S', 'D'],
 ['Z', 'S', 'Z', 'V', 'C'],
 ['O', 'D', 'K', 'V', 'X'],
 ['C', 'V', 'S', 'K', 'O'],
 ['H', 'C', 'D', 'V', 'O'],
 ['S', 'Z', 'H', 'D', 'S'],
 ['Z', 'S', 'S', 'Z', 'D'],
 ['Z', 'O', 'D', 'D', 'Z'],
 ['V', 'S', 'Z', 'D', 'S'],
 ['H', 'O', 'D', 'O', 'D']]

38 Gemini 2.5 Pro (low res)

Gemini

Threshold Row

56.4%

Mean Accuracy

11/12

Attempted Rows

24px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	V	Z	C	?	Z	D	V	Z	C	1/5 (20%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "V", "Z", "C"],
  [],
  [],
  [],
  []
]
```

39T Gemini 2.0 Flash (low res)

Gemini

Threshold Row

55.0%

Mean Accuracy

12/12

Attempted Rows

24px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	?	Z	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	D	V	?	?	?	N	V	D	N	S	1/5 (20%)
9	12px	Z	N	?	?	?	O	N	K	K	S	1/5 (20%)
10	10px	V	D	?	?	?	V	C	V	D	C	1/5 (20%)
11	8px	N	C	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	S	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's the breakdown of the letters in the image, presented as a list of lists, row by row:

```
[['S', 'D', 'C', 'D', 'Z'],
 ['C', 'D', 'D', 'O', 'K'],
 ['C', 'C', 'C', 'O', 'Z'],
 ['S', 'Z', 'S', 'H', 'D'],
 ['H', 'D', 'C', 'V', 'S'],
 ['K', 'D', 'V', 'S', 'D'],
 ['not sure', 'Z'],
 ['D', 'V'],
 ['Z', 'N'],
 ['V', 'D'],
 ['N', 'C'],
 ['S']]
```

40T Gemini 2.5 Flash (no thinking)

Gemini

Threshold Row

55.0%

Mean Accuracy

12/12

Attempted Rows

24px

Smallest Letters

Configuration: thinking_budget: 0, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	N	V	C	?	Z	D	V	Z	C	2/5 (40%)
8	15px	V	D	S	?	?	N	V	D	N	S	0/5 (0%)
9	12px	K	C	Z	H	N	O	N	K	K	S	0/5 (0%)
10	10px	?	Z	S	N	D	V	C	V	D	C	0/5 (0%)
11	8px	C	H	?	V	?	O	H	S	R	H	1/5 (20%)
12	6px	R	?	S	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here are the letters identified row by row:

[['S', 'D', 'C', 'D', 'Z'],
 ['C', 'D', 'D', 'O', 'K'],
 ['C', 'C', 'C', 'O', 'Z'],
 ['S', 'Z', 'S', 'H', 'D'],
 ['H', 'D', 'C', 'V', 'S'],
 ['K', 'D', 'V', 'S', 'D'],
 ['Z', 'N', 'V', 'C', 'not sure'],
 ['V', 'D', 'S', 'not sure', 'not sure'],
 ['K', 'C', 'Z', 'H', 'N'],
 ['not sure', 'Z', 'S', 'N', 'D'],
 ['C', 'H', 'not sure', 'V', 'not sure'],
 ['R', 'not sure', 'S', 'not sure', 'not sure'],
 ['D', 'C', 'D', 'D', 'D'],
 ['not sure', 'not sure', 'not sure', 'R', 'R'],
 ['V', 'not sure', 'C', 'K', 'O'],
 ['H', 'V', 'not sure', 'O', 'Z'],
 ['not sure', 'not sure', 'O', 'O', 'not sure'],
 ['C', 'not sure', 'D', 'C', 'D'],
 ['S', 'V', 'S', 'H', 'V'],
 ['C', 'Z', 'C', 'D', 'O']]

41T Gemini 2.5 Pro (medium res)

Gemini

Threshold Row

55.0%

Mean Accuracy

12/12

Attempted Rows

24px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	?	V	Z	?	Z	D	V	Z	C	3/5 (60%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "not sure", "V", "Z", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]
```

42 GPT-4O-MINI (high detail)

OpenAI

Threshold Row

51.7%

Mean Accuracy

12/12

Attempted Rows

24px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	O	V	S	D	K	D	V	S	D	4/5 (80%)
7	19px	Z	R	Y	Z	?	Z	D	V	Z	C	2/5 (40%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's the output representing the letters in the image, organized as a list of lists:

```plaintext
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'D', 'O', 'K'],
    ['C', 'C', 'C', 'O', 'Z'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['H', 'D', 'C', 'V', 'S'],
    ['K', 'O', 'V', 'S', 'D'],
    ['Z', 'R', 'Y', 'Z', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
    ['not sure', 'not sure', 'not sure', 'not sure', 'not sure'],
]
```

Note: The last rows contain "not sure" indicating uncertainty about the letters present.

43 GPT-4O-MINI (low detail)

OpenAI

Threshold Row

42.0%

Mean Accuracy

10/12

Attempted Rows

38px

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	O	Z	?	C	C	C	O	Z	2/5 (40%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	Y	?	H	D	C	V	S	3/5 (60%)
6	24px	*	D	C	Y	*	K	D	V	S	D	1/5 (20%)
7	19px	?	*	?	*	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here’s the output based on the image:

```plaintext
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'D', 'O', 'K'],
    ['C', 'C', 'O', 'Z', 'not sure'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['H', 'D', 'C', 'Y', 'not sure'],
    ['*', 'D', 'C', 'Y', '*'],
    ['.', '*', '.', '*'],
    ['.', '.', '.', '.'],
    ['.', '.', 'not sure', '.'],
    ['.', '.', '.', '.']
]
```

44 Grok 4 0709

xAI

Threshold Row

85.0%

Mean Accuracy

4/12

Attempted Rows

48px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	O	O	Z	C	C	C	O	Z	4/5 (80%)
4	38px	S	Z	S	W	O	S	Z	S	H	D	3/5 (60%)
5	30px	?	?	?	?	?	H	D	C	V	S	0/5 (0%)
6	24px	?	?	?	?	?	K	D	V	S	D	0/5 (0%)
7	19px	?	?	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[["S", "D", "C", "D", "Z"], ["C", "D", "D", "O", "K"], ["C", "C", "O", "O", "Z"], ["S", "Z", "S", "W", "O"]]

45T GPT-5.5 (none) (low detail)

OpenAI

Threshold Row

16.7%

Mean Accuracy

12/12

Attempted Rows

60px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	?	?	?	?	?	C	C	C	O	Z	0/5 (0%)
4	38px	?	?	?	?	?	S	Z	S	H	D	0/5 (0%)
5	30px	?	?	?	?	?	H	D	C	V	S	0/5 (0%)
6	24px	?	?	?	?	?	K	D	V	S	D	0/5 (0%)
7	19px	?	?	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]

46T GPT-5.6 (none) (low detail)

OpenAI

Threshold Row

16.7%

Mean Accuracy

12/12

Attempted Rows

60px

Smallest Letters

Configuration: reasoning_effort: none, image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	?	?	?	?	?	C	C	C	O	Z	0/5 (0%)
4	38px	A	?	?	?	?	S	Z	S	H	D	0/5 (0%)
5	30px	R	?	?	?	?	H	D	C	V	S	0/5 (0%)
6	24px	?	?	?	?	?	K	D	V	S	D	0/5 (0%)
7	19px	?	?	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["A", "not sure", "not sure", "not sure", "not sure"],
  ["R", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"],
  ["not sure", "not sure", "not sure", "not sure", "not sure"]
]

47 GPT-4O (low detail)

OpenAI

None

Threshold Row

0.0%

Mean Accuracy

0/12

Attempted Rows

N/A

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	?	?	?	?	?	S	D	C	D	Z	0/5 (0%)
2	60px	?	?	?	?	?	C	D	D	O	K	0/5 (0%)
3	48px	?	?	?	?	?	C	C	C	O	Z	0/5 (0%)
4	38px	?	?	?	?	?	S	Z	S	H	D	0/5 (0%)
5	30px	?	?	?	?	?	H	D	C	V	S	0/5 (0%)
6	24px	?	?	?	?	?	K	D	V	S	D	0/5 (0%)
7	19px	?	?	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

I'm unable to transcribe letters from this image format. Please provide a description or another format if possible.

Test Image (1K)

Click image to expand/collapse

Leaderboard (1K)

Rank	Model	Provider	Input Tokens	Output Tokens	Threshold Row	Threshold Size	Mean Accuracy	Attempted Rows
1T	Claude Haiku 4 5 20251001	Anthropic	1,419	251	12	6px	100.0%	12/12
2T	Claude Opus 4 5 20251101	Anthropic	1,419	227	12	6px	100.0%	12/12
3T	Claude Sonnet 4 20250514	Anthropic	1,419	227	12	6px	100.0%	12/12
4T	Claude Sonnet 4 5 20250929	Anthropic	1,419	275	12	6px	100.0%	12/12
5T	GPT-4.1	OpenAI	810	253	12	6px	100.0%	12/12
6T	GPT-4.1-MINI	OpenAI	1,705	195	12	6px	100.0%	12/12
7T	GPT-4.1-MINI (high detail)	OpenAI	1,705	195	12	6px	100.0%	12/12
8T	GPT-4.1-MINI (low detail)	OpenAI	1,705	195	12	6px	100.0%	12/12
9T	GPT-4.1-NANO (high detail)	OpenAI	2,566	195	12	6px	100.0%	12/12
10T	GPT-4.1-NANO (low detail)	OpenAI	2,566	195	12	6px	100.0%	12/12
11T	GPT-5 (minimal)	OpenAI	674	200	12	6px	100.0%	12/12
12T	GPT-5-MINI (minimal)	OpenAI	1,273	188	12	6px	100.0%	12/12
13T	GPT-5.4 (original detail)	OpenAI	1,273	198	12	6px	100.0%	12/12
14T	Gemini 2.5 Flash (no thinking)	Gemini	298	202	12	6px	100.0%	12/12
15T	Gemini 2.5 Flash (no thinking) (medium res)	Gemini	298	202	12	6px	100.0%	12/12
16T	Gemini 2.5 Pro	Gemini	298	212	12	6px	100.0%	12/12
17T	Gemini 2.5 Pro (medium res)	Gemini	298	212	12	6px	100.0%	12/12
18T	Gemini 3 Flash Preview (thinking: minimal) (high res)	Gemini	1,129	185	12	6px	100.0%	12/12
19T	Gemini 3 Flash Preview (thinking: minimal) (low res)	Gemini	296	242	12	6px	100.0%	12/12
20T	Gemini 3 Flash Preview (thinking: minimal) (medium res)	Gemini	569	242	12	6px	100.0%	12/12
21T	Gemini 3 Pro Preview (high res)	Gemini	1,129	212	12	6px	100.0%	12/12
22T	Gemini 3 Pro Preview (low res)	Gemini	296	232	12	6px	100.0%	12/12
23T	Gemini 3 Pro Preview (medium res)	Gemini	569	232	12	6px	100.0%	12/12
24T	Claude Opus 4 1 20250805	Anthropic	1,419	230	12	6px	98.3%	12/12
25T	Gemini 2.0 Flash (medium res)	Gemini	297	255	12	6px	98.3%	12/12
26	GPT-4O (high detail)	OpenAI	810	193	12	6px	68.3%	12/12
27	Gemini 2.0 Flash (high res)	Gemini	1,329	182	10	10px	100.0%	10/12
28	Grok 4 0709	xAI	978	191	9	12px	75.0%	12/12
29	Gemini 2.0 Flash (low res)	Gemini	105	208	2	60px	23.6%	11/12
30	Gemini 2.5 Pro (low res)	Gemini	106	196	1	76px	30.9%	11/12
31	GPT-4O-MINI (high detail)	OpenAI	25,546	199	1	76px	23.6%	11/12
32	GPT-4O-MINI (low detail)	OpenAI	2,878	186	1	76px	18.0%	10/12
33	GPT-4O (low detail)	OpenAI	130	40	None	N/A	0.0%	0/12

Click a row to see detailed results
Use ← → arrow keys to navigate between models when a card is open

1T Claude Haiku 4 5 20251001

Anthropic

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

# Row by Row Letter Analysis

Here's the list of lists representing the letters:

```
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

All letters are clearly visible and identified with confidence. No "not sure" entries needed.

2T Claude Opus 4 5 20251101

Anthropic

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

I'll read the letters row by row from the image:

```
[
    ["S", "D", "C", "D", "Z"],
    ["C", "D", "D", "O", "K"],
    ["C", "C", "C", "O", "Z"],
    ["S", "Z", "S", "H", "D"],
    ["H", "D", "C", "V", "S"],
    ["K", "D", "V", "S", "D"],
    ["Z", "D", "V", "Z", "C"],
    ["N", "V", "D", "N", "S"],
    ["O", "N", "K", "K", "S"],
    ["V", "C", "V", "D", "C"],
    ["O", "H", "S", "R", "H"],
    ["C", "D", "Z", "S", "H"]
]
```

3T Claude Sonnet 4 20250514

Anthropic

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Looking at the image row by row, here are the letters I can identify:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

4T Claude Sonnet 4 5 20250929

Anthropic

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

I'll read the letters row by row from the image:

```
[
  ['S', 'D', 'C', 'D', 'Z'],
  ['C', 'D', 'D', 'O', 'K'],
  ['C', 'C', 'C', 'O', 'Z'],
  ['S', 'Z', 'S', 'H', 'D'],
  ['H', 'D', 'C', 'V', 'S'],
  ['K', 'D', 'V', 'S', 'D'],
  ['Z', 'D', 'V', 'Z', 'C'],
  ['N', 'V', 'D', 'N', 'S'],
  ['O', 'N', 'K', 'K', 'S'],
  ['V', 'C', 'V', 'D', 'C'],
  ['O', 'H', 'S', 'R', 'H'],
  ['C', 'D', 'Z', 'S', 'H']
]
```

5T GPT-4.1

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Here are the letters from the image, row by row, represented as a list of lists. For any letter I am not sure of, I have marked it as "not sure":

```python
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'D', 'O', 'K'],
    ['C', 'C', 'C', 'O', 'Z'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['H', 'D', 'C', 'V', 'S'],
    ['K', 'D', 'V', 'S', 'D'],
    ['z', 'D', 'V', 'z', 'C'],
    ['N', 'V', 'D', 'N', 'S'],
    ['O', 'N', 'K', 'K', 'S'],
    ['v', 'C', 'V', 'D', 'C'],
    ['O', 'H', 'S', 'R', 'H'],
    ['C', 'D', 'Z', 'S', 'H']
]
```

Let me know if you need the output in a different format or more details!

6T GPT-4.1-MINI

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

7T GPT-4.1-MINI (high detail)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

8T GPT-4.1-MINI (low detail)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

9T GPT-4.1-NANO (high detail)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ['s', 'D', 'C', 'D', 'Z'],
  ['c', 'D', 'D', 'o', 'K'],
  ['c', 'c', 'c', 'o', 'z'],
  ['s', 'z', 's', 'H', 'D'],
  ['H', 'D', 'c', 'v', 's'],
  ['K', 'D', 'v', 's', 'D'],
  ['z', 'D', 'v', 'z', 'c'],
  ['N', 'v', 'D', 'N', 's'],
  ['o', 'N', 'k', 'K', 's'],
  ['v', 'c', 'v', 'D', 'c'],
  ['o', 'H', 's', 'R', 'H'],
  ['c', 'D', 'z', 's', 'H']
]

10T GPT-4.1-NANO (low detail)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ['s', 'D', 'C', 'D', 'Z'],
  ['c', 'D', 'D', 'o', 'K'],
  ['c', 'c', 'c', 'o', 'z'],
  ['s', 'z', 's', 'H', 'D'],
  ['H', 'D', 'c', 'v', 's'],
  ['K', 'D', 'v', 's', 'D'],
  ['z', 'D', 'v', 'z', 'c'],
  ['N', 'v', 'D', 'N', 's'],
  ['o', 'N', 'k', 'K', 's'],
  ['v', 'c', 'v', 'D', 'c'],
  ['o', 'H', 's', 'R', 'H'],
  ['c', 'D', 'z', 's', 'H']
]

11T GPT-5 (minimal)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: minimal, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

12T GPT-5-MINI (minimal)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: reasoning_effort: minimal, image_detail: auto, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
["S", "D", "C", "D", "Z"],
["C", "D", "D", "O", "K"],
["C", "C", "C", "O", "Z"],
["S", "Z", "S", "H", "D"],
["H", "D", "C", "V", "S"],
["K", "D", "V", "S", "D"],
["Z", "D", "V", "Z", "C"],
["N", "V", "D", "N", "S"],
["O", "N", "K", "K", "S"],
["V", "C", "V", "D", "C"],
["O", "H", "S", "R", "H"],
["C", "D", "Z", "S", "H"]
]

13T GPT-5.4 (original detail)

OpenAI

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: original, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]

14T Gemini 2.5 Flash (no thinking)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: thinking_budget: 0, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Here are the letters read out row by row:

[['S', 'D', 'C', 'D', 'Z'],
 ['C', 'D', 'D', 'O', 'K'],
 ['C', 'C', 'C', 'O', 'Z'],
 ['S', 'Z', 'S', 'H', 'D'],
 ['H', 'D', 'C', 'V', 'S'],
 ['K', 'D', 'V', 'S', 'D'],
 ['Z', 'D', 'V', 'Z', 'C'],
 ['N', 'V', 'D', 'N', 'S'],
 ['O', 'N', 'K', 'K', 'S'],
 ['V', 'C', 'V', 'D', 'C'],
 ['O', 'H', 'S', 'R', 'H'],
 ['C', 'D', 'Z', 'S', 'H']]

15T Gemini 2.5 Flash (no thinking) (medium res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: thinking_budget: 0, media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Here are the letters, read row by row:

[['S', 'D', 'C', 'D', 'Z'],
 ['C', 'D', 'D', 'O', 'K'],
 ['C', 'C', 'C', 'O', 'Z'],
 ['S', 'Z', 'S', 'H', 'D'],
 ['H', 'D', 'C', 'V', 'S'],
 ['K', 'D', 'V', 'S', 'D'],
 ['Z', 'D', 'V', 'Z', 'C'],
 ['N', 'V', 'D', 'N', 'S'],
 ['O', 'N', 'K', 'K', 'S'],
 ['V', 'C', 'V', 'D', 'C'],
 ['O', 'H', 'S', 'R', 'H'],
 ['C', 'D', 'Z', 'S', 'H']]

16T Gemini 2.5 Pro

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

17T Gemini 2.5 Pro (medium res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

18T Gemini 3 Flash Preview (thinking: minimal) (high res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: HIGH, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Based on the image provided, here are the letters row by row:

Row 1: S, D, C, D, Z
Row 2: C, D, D, O, K
Row 3: C, C, C, O, Z
Row 4: S, Z, S, H, D
Row 5: H, D, C, V, S
Row 6: K, D, V, S, D
Row 7: Z, D, V, Z, C
Row 8: N, V, D, N, S
Row 9: O, N, K, K, S
Row 10: V, C, V, D, C
Row 11: O, H, S, R, H
Row 12: C, D, Z, S, H

19T Gemini 3 Flash Preview (thinking: minimal) (low res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Based on the image provided, here are the letters row by row:

1.  ['S', 'D', 'C', 'D', 'Z']
2.  ['C', 'D', 'D', 'O', 'K']
3.  ['C', 'C', 'C', 'O', 'Z']
4.  ['S', 'Z', 'S', 'H', 'D']
5.  ['H', 'D', 'C', 'V', 'S']
6.  ['K', 'D', 'V', 'S', 'D']
7.  ['Z', 'D', 'V', 'Z', 'C']
8.  ['N', 'V', 'D', 'N', 'S']
9.  ['O', 'N', 'K', 'K', 'S']
10. ['V', 'C', 'V', 'D', 'C']
11. ['O', 'H', 'S', 'R', 'H']
12. ['C', 'D', 'Z', 'S', 'H']

20T Gemini 3 Flash Preview (thinking: minimal) (medium res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Based on the image provided, here are the letters row by row:

1.  ['S', 'D', 'C', 'D', 'Z']
2.  ['C', 'D', 'D', 'O', 'K']
3.  ['C', 'C', 'C', 'O', 'Z']
4.  ['S', 'Z', 'S', 'H', 'D']
5.  ['H', 'D', 'C', 'V', 'S']
6.  ['K', 'D', 'V', 'S', 'D']
7.  ['Z', 'D', 'V', 'Z', 'C']
8.  ['N', 'V', 'D', 'N', 'S']
9.  ['O', 'N', 'K', 'K', 'S']
10. ['V', 'C', 'V', 'D', 'C']
11. ['O', 'H', 'S', 'R', 'H']
12. ['C', 'D', 'Z', 'S', 'H']

21T Gemini 3 Pro Preview (high res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: HIGH, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

22T Gemini 3 Pro Preview (low res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Based on the image provided, here is the row-by-row transcription of the letters:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

23T Gemini 3 Pro Preview (medium res)

Gemini

Threshold Row

100.0%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Based on the image provided, here is the list of lists representing the letters in each row:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "R", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

24T Claude Opus 4 1 20250805

Anthropic

Threshold Row

98.3%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 2000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	H	H	O	H	S	R	H	4/5 (80%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Looking at the image row by row, I can identify the following letters:

```
[
  ["S", "D", "C", "D", "Z"],
  ["C", "D", "D", "O", "K"],
  ["C", "C", "C", "O", "Z"],
  ["S", "Z", "S", "H", "D"],
  ["H", "D", "C", "V", "S"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "V", "D", "C"],
  ["O", "H", "S", "H", "H"],
  ["C", "D", "Z", "S", "H"]
]
```

25T Gemini 2.0 Flash (medium res)

Gemini

Threshold Row

98.3%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: media_resolution: MEDIUM, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	G	V	C	V	D	C	4/5 (80%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

Here's the letter data extracted from the image, row by row:

*   `['S', 'D', 'C', 'D', 'Z']`
*   `['C', 'D', 'D', 'O', 'K']`
*   `['C', 'C', 'C', 'O', 'Z']`
*   `['S', 'Z', 'S', 'H', 'D']`
*   `['H', 'D', 'C', 'V', 'S']`
*   `['K', 'D', 'V', 'S', 'D']`
*   `['Z', 'D', 'V', 'Z', 'C']`
*   `['N', 'V', 'D', 'N', 'S']`
*   `['O', 'N', 'K', 'K', 'S']`
*   `['V', 'C', 'V', 'D', 'G']`
*   `['O', 'H', 'S', 'R', 'H']`
*   `['C', 'D', 'Z', 'S', 'H']`

26 GPT-4O (high detail)

OpenAI

Threshold Row

68.3%

Mean Accuracy

12/12

Attempted Rows

6px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	O	C	K	C	D	D	O	K	3/5 (60%)
3	48px	C	C	O	Z	?	C	C	C	O	Z	2/5 (40%)
4	38px	S	Z	H	D	S	S	Z	S	H	D	2/5 (40%)
5	30px	H	D	V	S	D	H	D	C	V	S	2/5 (40%)
6	24px	K	D	V	Z	C	K	D	V	S	D	3/5 (60%)
7	19px	Z	V	D	N	S	Z	D	V	Z	C	1/5 (20%)
8	15px	N	V	D	S	?	N	V	D	N	S	3/5 (60%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	O	H	S	R	H	O	H	S	R	H	5/5 (100%)
12	6px	C	D	Z	S	H	C	D	Z	S	H	5/5 (100%)

Full Response:

```python
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'D', 'O', 'C', 'K'],
    ['C', 'C', 'O', 'Z'],
    ['S', 'Z', 'H', 'D', 'S'],
    ['H', 'D', 'V', 'S', 'D'],
    ['K', 'D', 'V', 'Z', 'C'],
    ['Z', 'V', 'D', 'N', 'S'],
    ['N', 'V', 'D', 'S'],
    ['O', 'N', 'K', 'K', 'S'],
    ['V', 'C', 'V', 'D', 'C'],
    ['O', 'H', 'S', 'R', 'H'],
    ['C', 'D', 'Z', 'S', 'H']
]
```

27 Gemini 2.0 Flash (high res)

Gemini

Threshold Row

100.0%

Mean Accuracy

10/12

Attempted Rows

10px

Smallest Letters

Configuration: media_resolution: HIGH, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	V	C	V	D	C	V	C	V	D	C	5/5 (100%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's the list of lists representing the letters in the image, row by row:

```
[['S', 'D', 'C', 'D', 'Z'],
['C', 'D', 'D', 'O', 'K'],
['C', 'C', 'C', 'O', 'Z'],
['S', 'Z', 'S', 'H', 'D'],
['H', 'D', 'C', 'V', 'S'],
['K', 'D', 'V', 'S', 'D'],
['Z', 'D', 'V', 'Z', 'C'],
['N', 'V', 'D', 'N', 'S'],
['O', 'N', 'K', 'K', 'S'],
['V', 'C', 'V', 'D', 'C']]
```

28 Grok 4 0709

xAI

Threshold Row

75.0%

Mean Accuracy

12/12

Attempted Rows

12px

Smallest Letters

Configuration: temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	D	D	O	K	C	D	D	O	K	5/5 (100%)
3	48px	C	C	C	O	Z	C	C	C	O	Z	5/5 (100%)
4	38px	S	Z	S	H	D	S	Z	S	H	D	5/5 (100%)
5	30px	H	D	C	V	S	H	D	C	V	S	5/5 (100%)
6	24px	K	D	V	S	D	K	D	V	S	D	5/5 (100%)
7	19px	Z	D	V	Z	C	Z	D	V	Z	C	5/5 (100%)
8	15px	N	V	D	N	S	N	V	D	N	S	5/5 (100%)
9	12px	O	N	K	K	S	O	N	K	K	S	5/5 (100%)
10	10px	C	V	D	C	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

[["S", "D", "C", "D", "Z"],
["C", "D", "D", "O", "K"],
["C", "C", "C", "O", "Z"],
["S", "Z", "S", "H", "D"],
["H", "D", "C", "V", "S"],
["K", "D", "V", "S", "D"],
["Z", "D", "V", "Z", "C"],
["N", "V", "D", "N", "S"],
["O", "N", "K", "K", "S"],
["C", "V", "D", "C", "not sure"],
["not sure", "not sure", "not sure", "not sure", "not sure"],
["not sure", "not sure", "not sure", "not sure", "not sure"]]

29 Gemini 2.0 Flash (low res)

Gemini

Threshold Row

23.6%

Mean Accuracy

11/12

Attempted Rows

60px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	C	D	O	K	C	D	D	O	K	4/5 (80%)
3	48px	S	Z	S	H	D	C	C	C	O	Z	0/5 (0%)
4	38px	H	D	C	V	S	S	Z	S	H	D	0/5 (0%)
5	30px	K	D	V	S	D	H	D	C	V	S	1/5 (20%)
6	24px	Z	D	V	Z	C	K	D	V	S	D	2/5 (40%)
7	19px	N	V	D	N	S	Z	D	V	Z	C	0/5 (0%)
8	15px	O	N	K	K	S	N	V	D	N	S	1/5 (20%)
9	12px	V	C	V	D	C	O	N	K	K	S	0/5 (0%)
10	10px	O	H	S	R	H	V	C	V	D	C	0/5 (0%)
11	8px	C	D	Z	S	C	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's the OCR output in the format you requested:

```
[
  ['S', 'D', 'C', 'D', 'Z'],
  ['C', 'C', 'D', 'O', 'K'],
  ['S', 'Z', 'S', 'H', 'D'],
  ['H', 'D', 'C', 'V', 'S'],
  ['K', 'D', 'V', 'S', 'D'],
  ['Z', 'D', 'V', 'Z', 'C'],
  ['N', 'V', 'D', 'N', 'S'],
  ['O', 'N', 'K', 'K', 'S'],
  ['V', 'C', 'V', 'D', 'C'],
  ['O', 'H', 'S', 'R', 'H'],
  ['C', 'D', 'Z', 'S', 'C']
]
```

30 Gemini 2.5 Pro (low res)

Gemini

Threshold Row

30.9%

Mean Accuracy

11/12

Attempted Rows

76px

Smallest Letters

Configuration: media_resolution: LOW, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	C	O	O	K	C	D	D	O	K	3/5 (60%)
3	48px	S	Z	C	H	S	C	C	C	O	Z	1/5 (20%)
4	38px	H	D	C	V	D	S	Z	S	H	D	1/5 (20%)
5	30px	K	D	V	S	D	H	D	C	V	S	1/5 (20%)
6	24px	Z	D	V	Z	C	K	D	V	S	D	2/5 (40%)
7	19px	N	V	D	N	S	Z	D	V	Z	C	0/5 (0%)
8	15px	O	N	K	K	S	N	V	D	N	S	1/5 (20%)
9	12px	V	C	S	R	H	O	N	K	K	S	0/5 (0%)
10	10px	O	H	V	D	?	V	C	V	D	C	2/5 (40%)
11	8px	C	D	Z	S	H	O	H	S	R	H	1/5 (20%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

```json
[
  ["S", "D", "C", "D", "Z"],
  ["C", "C", "O", "O", "K"],
  ["S", "Z", "C", "H", "S"],
  ["H", "D", "C", "V", "D"],
  ["K", "D", "V", "S", "D"],
  ["Z", "D", "V", "Z", "C"],
  ["N", "V", "D", "N", "S"],
  ["O", "N", "K", "K", "S"],
  ["V", "C", "S", "R", "H"],
  ["O", "H", "V", "D", "not sure"],
  ["C", "D", "Z", "S", "H"]
]
```

31 GPT-4O-MINI (high detail)

OpenAI

Threshold Row

23.6%

Mean Accuracy

11/12

Attempted Rows

76px

Smallest Letters

Configuration: image_detail: high, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	C	C	O	K	C	D	D	O	K	3/5 (60%)
3	48px	S	Z	S	H	D	C	C	C	O	Z	0/5 (0%)
4	38px	H	D	C	V	S	S	Z	S	H	D	0/5 (0%)
5	30px	K	D	V	S	D	H	D	C	V	S	1/5 (20%)
6	24px	Z	D	V	Z	C	K	D	V	S	D	2/5 (40%)
7	19px	N	V	D	N	S	Z	D	V	Z	C	0/5 (0%)
8	15px	O	N	K	K	S	N	V	D	N	S	1/5 (20%)
9	12px	V	C	V	D	C	O	N	K	K	S	0/5 (0%)
10	10px	O	H	S	R	H	V	C	V	D	C	0/5 (0%)
11	8px	C	D	Z	S	H	O	H	S	R	H	1/5 (20%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here is a list of letters organized row by row as observed in the image:

```plaintext
[
  ['S', 'D', 'C', 'D', 'Z'],
  ['C', 'C', 'C', 'O', 'K'],
  ['S', 'Z', 'S', 'H', 'D'],
  ['H', 'D', 'C', 'V', 'S'],
  ['K', 'D', 'V', 'S', 'D'],
  ['z', 'D', 'V', 'z', 'C'],
  ['N', 'V', 'D', 'N', 'S'],
  ['O', 'N', 'K', 'K', 'S'],
  ['V', 'C', 'V', 'D', 'C'],
  ['O', 'H', 'S', 'R', 'H'],
  ['C', 'D', 'Z', 'S', 'H']
]
```

32 GPT-4O-MINI (low detail)

OpenAI

Threshold Row

18.0%

Mean Accuracy

10/12

Attempted Rows

76px

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	S	D	C	D	Z	S	D	C	D	Z	5/5 (100%)
2	60px	C	C	C	O	Z	C	D	D	O	K	2/5 (40%)
3	48px	S	Z	S	H	D	C	C	C	O	Z	0/5 (0%)
4	38px	K	D	V	S	D	S	Z	S	H	D	1/5 (20%)
5	30px	Z	D	V	Z	C	H	D	C	V	S	1/5 (20%)
6	24px	N	V	D	N	S	K	D	V	S	D	0/5 (0%)
7	19px	O	N	K	K	S	Z	D	V	Z	C	0/5 (0%)
8	15px	V	O	V	D	C	N	V	D	N	S	0/5 (0%)
9	12px	?	S	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	C	D	Z	S	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

Here's the representation of the letters from the image, row by row:

```plaintext
[
    ['S', 'D', 'C', 'D', 'Z'],
    ['C', 'C', 'C', 'O', 'Z'],
    ['S', 'Z', 'S', 'H', 'D'],
    ['K', 'D', 'V', 'S', 'D'],
    ['Z', 'D', 'V', 'Z', 'C'],
    ['N', 'V', 'D', 'N', 'S'],
    ['O', 'N', 'K', 'K', 'S'],
    ['V', 'O', 'V', 'D', 'C'],
    ['not sure', 'S', 'not sure', 'not sure', 'not sure'],
    ['C', 'D', 'Z', 'S', 'not sure']
]
```

33 GPT-4O (low detail)

OpenAI

None

Threshold Row

0.0%

Mean Accuracy

0/12

Attempted Rows

N/A

Smallest Letters

Configuration: image_detail: low, temperature: 1.0, max_tokens: 20000

Row	Size	Predicted					Ground Truth					Accuracy
1	76px	?	?	?	?	?	S	D	C	D	Z	0/5 (0%)
2	60px	?	?	?	?	?	C	D	D	O	K	0/5 (0%)
3	48px	?	?	?	?	?	C	C	C	O	Z	0/5 (0%)
4	38px	?	?	?	?	?	S	Z	S	H	D	0/5 (0%)
5	30px	?	?	?	?	?	H	D	C	V	S	0/5 (0%)
6	24px	?	?	?	?	?	K	D	V	S	D	0/5 (0%)
7	19px	?	?	?	?	?	Z	D	V	Z	C	0/5 (0%)
8	15px	?	?	?	?	?	N	V	D	N	S	0/5 (0%)
9	12px	?	?	?	?	?	O	N	K	K	S	0/5 (0%)
10	10px	?	?	?	?	?	V	C	V	D	C	0/5 (0%)
11	8px	?	?	?	?	?	O	H	S	R	H	0/5 (0%)
12	6px	?	?	?	?	?	C	D	Z	S	H	0/5 (0%)

Full Response:

I'm unable to view the image directly, but I can help with analyzing it if you can provide a description or transcribe the visible letters for me. Let me know how you'd like to proceed!