BigCode Models Leaderboard — leaderboard
Comprehensive 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. Status: saturated. 60 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Nxcode-CQ-7B-orpo | 87.2 |
| 2 | Qwen 2.5 Coder 32B Instruct | 83.2 |
| 3 | WizardCoder-Python-34B-V1.0 | 70.7 |
| 4 | Qwen 2.5 Coder 32B | 57.1 |
| 5 | deepseek-coder-33B-base | 52.5 |
| 6 | Phi-1 | 51.2 |
| 7 | CodeQwen1.5-7B | 50.8 |
| 8 | CodeLlama-34B-Instruct | 50.8 |
| 9 | starcoder2-15B | 44.1 |
| 10 | falcon-180B | 35.4 |
| 11 | starcoder2-7B | 34.1 |
| 12 | starcoder2-3B | 31.4 |
Interactive version: theaggregate.ai/benchmark?slug=bigcode-models-leaderboard · How the rankings work · Data refreshed daily, snapshot 2026-07-22.