BigCode Models Leaderboard: leaderboard
Leaderboard evaluating code generation models on HumanEval, MultiPL-E, DS-1000, and related benchmarks. Covers 60+ models including StarCoder, CodeLlama, Qwen-Coder, and DeepSeek-Coder variants.
Metric: HumanEval Python Pass@1 (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 60 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nxcode-CQ-7B-orpo | 87.2 | #1477 |
| 2 | CodeQwen1.5-7B Chat | 87.2 | #1144 |
| 3 | Qwen 2.5 Coder 32B Instruct | 83.2 | #528 |
| 4 | Phind-CodeLlama-34B-v2 | 72 | #1589 |
| 5 | WizardCoder-15B-V1.0 | 58.1 | #1588 |
| 6 | Qwen 2.5 Coder 32B | 57.1 | #667 |
| 7 | Phi-1 | 51.2 | #1639 |
| 8 | CodeQwen1.5-7B | 50.8 | #1514 |
| 9 | starcoder2-15B | 44.1 | #1495 |
| 10 | starcoder2-7B | 34.1 | #1572 |
| 11 | starcoder2-3B | 31.4 | #1596 |
| 12 | Stable-Code 3B | 30.7 | #1595 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=bigcode-models-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.