McEval: leaderboard
McEval: Measures model capability on programming, code generation, code repair, or repository-level software tasks.
Metric: AVG (%). Source: github.com. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4 Turbo | 63.4 | #404 |
| 2 | GPT-3.5 Turbo | 52.6 | #865 |
| 3 | Codestral-22B-v0.1 | 50.5 | #911 |
| 4 | Magicoder-S-DS-6.7B | 48.6 | #1256 |
| 5 | CodeQwen1.5-7B | 45.5 | #1514 |
| 6 | Nxcode-CQ-7B-orpo | 44.7 | #1477 |
| 7 | Llama 3 8B Instruct | 36 | #1129 |
| 8 | Qwen 1.5 72B Chat | 35.8 | #776 |
| 9 | Phi-3-medium-4k-instruct | 35.2 | #952 |
| 10 | WizardCoder-15B-V1.0 | 28 | #1588 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mceval · How It Works · Data refreshed daily, snapshot 2026-10-20.