InfiBench — leaderboard
Metric: Score (%). Source: infi-coder.github.io. 106 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 (0613) | 70.64 |
| 2 | GPT-4 Turbo (Preview) | 68.42 |
| 3 | GPT-4o (2024-05-13) | 66.19 |
| 4 | Claude 3 Opus | 63.89 |
| 5 | Mistral Large | 58.22 |
| 6 | Claude 3 Sonnet | 58.2 |
| 7 | Claude 3 Haiku | 57.57 |
| 8 | deepseek-llm-67B-chat | 57.41 |
| 9 | GPT-3.5 Turbo (0613) | 56.47 |
| 10 | Mistral Small | 55.62 |
| 11 | Mixtral 8x7B Instruct | 55.55 |
| 12 | deepseek-coder-6.7B-instruct | 53.25 |
| 13 | WizardCoder-Python-34B-V1.0 | 52.59 |
| 14 | CodeLlama-34B-Instruct | 50.45 |
| 15 | Yi 34B (Chat) | 49.58 |
Interactive version: theaggregate.ai/benchmark?slug=infibench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.