EvoEval Difficult — leaderboard
Metric: Pass@1 (%). Source: evo-eval.github.io. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 | 52 |
| 2 | GPT-4 Turbo | 50 |
| 3 | Claude 3 Haiku | 40 |
| 4 | deepseek-coder-6.7B-instruct | 40 |
| 5 | Gemini 1.0 Pro | 37 |
| 6 | Claude 2 | 29 |
| 7 | Mixtral 8x7B Instruct | 21 |
| 8 | deepseek-coder-1.3B-instruct | 20 |
| 9 | Phi-2 | 18 |
| 10 | starcoder2-15B | 16 |
| 11 | gemma-7B | 12 |
| 12 | starcoder | 12 |
| 13 | starcoder2-7B | 12 |
| 14 | Mistral 7B | 8 |
| 15 | starcoder2-3B | 8 |
Interactive version: aibenchmarks.dev/benchmark?slug=evoeval-difficult · How the rankings work · Data refreshed daily, snapshot 2026-07-08.