OCR benchmark
Higher is better
| # | Model | Score | Source |
|---|---|---|---|
| ★ | o4-mini (high) | 99.3 | codesota-api |
| 2 | o3-mini (high) | 97.6 | codesota-api |
| 3 | o4-mini | 97.3 | codesota-api |
| 4 | o3-mini | 96.3 | codesota-api |
| 5 | gpt-41 | 94.5 | codesota-api |
| 6 | GPT-4.1 mini | 93.8 | codesota-api |
| 7 | Qwen2.5-Coder-32B-Instruct | 92.7 | codesota-api |
| 8 | o1-preview | 92.4 | codesota-api |
| 9 | o1-mini | 92.4 | codesota-api |
| 10 | Claude 3.5 Sonnet (Oct 2024) | 92.1 | codesota-api |
| 11 | claude-35-sonnet | 92 | codesota-api |
| 12 | gpt-4o | 91 | codesota-api |
| 13 | GPT-4o (Nov 2024) | 90.2 | codesota-api |
| 14 | llama-31-405b | 89 | codesota-api |
| 15 | gpt-45-preview | 88.6 | codesota-api |
| 16 | grok-2 | 88.4 | codesota-api |
| 17 | Qwen2.5-Coder-7B-Instruct | 88.4 | codesota-api |
| 18 | o3 (high) | 88.4 | codesota-api |
| 19 | gpt-4-turbo | 88.2 | codesota-api |
| 20 | Gemma 3 27B IT | 87.8 | codesota-api |
| 21 | o3 | 87.4 | codesota-api |
| 22 | gpt-4o-mini | 87.2 | codesota-api |
| 23 | GPT-4.1 nano | 87 | codesota-api |
| 24 | Gemma 3 12B IT | 85.4 | codesota-api |
| 25 | DeepSeek-Coder-V2-Instruct | 85.4 | codesota-api |
| 26 | claude-3-opus | 84.9 | codesota-api |
| 27 | Phi-4 (14B) | 82.6 | codesota-api |
| 28 | deepseek-v3 | 82.6 | codesota-api |
| 29 | llama-3-70b | 81.7 | codesota-api |
| 30 | llama-31-70b | 80.5 | codesota-api |
| 31 | gemini-15-pro | 71.9 | codesota-api |
| 32 | Gemma 3 4B IT | 71.3 | codesota-api |
| 33 | DeepSeek-V3 | 65.2 | codesota-api |