BigCodeBench-Hard: leaderboard
BigCodeBench-Hard evaluates code generation on the harder BigCodeBench subset, reporting pass@1 in complete and instruct settings.
Metric: Instruct pass@1 (imported). Source: huggingface.co. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 Mini (2025-01-31) | 33.1 | #240 |
| 2 | Claude 3.7 Sonnet (20250219) | 32.4 | #201 |
| 3 | O1 (2024-12-17) | 32.4 | #149 |
| 4 | GPT-4.1 Mini | 31.8 | #355 |
| 5 | GPT-4.1 | 31.8 | #248 |
| 6 | Grok 3 Mini Beta | 31.1 | #296 |
| 7 | DeepSeek R1 | 29.7 | #251 |
| 8 | Gemini 2.5 Pro (Preview 03-25) | 29.7 | #123 |
| 9 | GPT-4 Turbo | 29.1 | #404 |
| 10 | Qwen 2.5 Max | 29.1 | #416 |
| 11 | Llama 3.3 70B Instruct | 28.4 | #534 |
| 12 | GPT-4.1 Nano | 28.4 | #731 |
| 13 | DeepSeek V3 | 28.4 | #321 |
| 14 | GPT-4o (2024-11-20) | 27.7 | #381 |
| 15 | O1 Mini (2024-09-12) | 27.7 | #403 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=bigcodebench-hard · How It Works · Data refreshed daily, snapshot 2026-10-20.