BigCodeBench: leaderboard
Coding benchmark with 1,140 function-level tasks requiring composition of multiple function calls and complex instructions.
Metric: Pass@1 (%). Source: bigcode-bench.github.io. Saturation forecast: Estimated already saturated. 126 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-05-13) | 51.1 | #325 |
| 2 | DeepSeek V3 | 50 | #321 |
| 3 | Llama 4 Maverick | 49.7 | #464 |
| 4 | Qwen 2.5 Coder 32B Instruct | 49 | #528 |
| 5 | GPT-4.1 Mini | 48.9 | #355 |
| 6 | DeepSeek V2 Chat (0628) | 48.9 | #400 |
| 7 | GPT-4 Turbo | 48.2 | #404 |
| 8 | Qwen 2.5 Coder 14B Instruct | 48.2 | #675 |
| 9 | GPT-4o (2024-11-20) | 48 | #381 |
| 10 | Llama 3.3 70B Instruct | 46.9 | #534 |
| 11 | Claude 3.5 Sonnet (20240620) | 46.8 | #299 |
| 12 | Llama 3.1 70B Instruct | 46.1 | #561 |
| 13 | GPT-4o Mini (2024-07-18) | 46.1 | #558 |
| 14 | Claude 3.5 Haiku (20241022) | 46.1 | #588 |
| 15 | GPT-4 (0613) | 46 | #491 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=bigcodebench · How It Works · Data refreshed daily, snapshot 2026-10-20.