EDIT-Bench Core Easy
Code · 2025-11-06
The EDIT-Bench Core easy slice reports pass@1 on the 62 distinct tasks classified as easy by the benchmark authors.
Top models (higher is better)
| Model | Score |
|---|---|
| Sonnet 4 | 96.8 |
| o3-mini | 93.5 |
| Sonnet 4.5 | 91.9 |
| Claude 3.7 Sonnet | 91.9 |
| Claude 3.5 Sonnet (Oct 2024) | 90.3 |
| O4 Mini | 90.3 |
| Kimi K2 Instruct 0905 | 88.7 |
| GPT-5 | 82.3 |
| Grok 4 Fast | 82.3 |
| DeepSeek-V3.1 | 80.7 |
| Grok Code Fast 1 | 80.7 |
| Llama 3.3 70B Instruct | 80.7 |
| Qwen2.5 72B Instruct | 80.7 |
| Qwen3-Coder-480B-A35B-Instruct | 80.7 |
| Gemini 2.5 Pro | 79.0 |
| GLM 4.6 | 79.0 |
| GPT-4o Mini | 79.0 |
| GPT-5 Mini | 79.0 |
| Qwen2.5 Coder 32B Instruct | 79.0 |
| Qwen3 14B | 79.0 |
| Devstral Medium 1.0 | 77.4 |
| gpt-oss-20b | 77.4 |
| GPT-4o (2024-08-06) | 75.8 |
| Llama 3.1 405B Instruct | 75.8 |
| Gemini 2.5 Flash | 74.2 |
| Llama 4 Maverick Instruct | 74.2 |
| Devstral Small 1.1 | 72.6 |
| GPT-5 Nano | 72.6 |
| Qwen3-30B-A3B | 72.6 |
| gpt-oss-120b | 71.0 |
| Qwen3 Coder Flash (rolling alias) | 71.0 |
| Llama 4 Scout Instruct | 69.3 |
| Mistral Small 3.2 24B Instruct 2506 | 66.1 |
| Codestral 25.08 | 64.5 |
| DeepSeek-R1-0528 | 62.9 |
| Llama 3.1 8B Instruct | 56.5 |
| Kimi-Dev-72B | 54.8 |
| Gemma 3n E4B Instruct | 51.6 |
| Gemma 3 27B IT | 45.2 |
| GLM-4.5 | 45.2 |