CodeQwen1.5-7B: benchmark results
Alibaba's CodeQwen1.5-7B, the code-specific base version of Qwen1.5 pretrained on 3 trillion tokens of code, supporting 92 programming languages with a 64K-token context. Provider: Alibaba. Released 2024-04-16. Access: Open.
Aggregate Rating 1275 ± 24, rank #1514 of 1656 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ShaderMatch | 16.75 | Clone Match Rate (%) | 90.5 |
| BigCode Models Leaderboard | 50.8 | HumanEval Python Pass@1 (%) | 60.2 |
| Big Code Memorization - HumanEval pass@50 | 38.84 | HumanEval pass@50 (%) | 58.8 |
| Big Code Memorization - HumanEval-ET pass@50 | 32.79 | HumanEval-ET pass@50 (%) | 58.8 |
| McEval | 45.5 | AVG (%) | 57.1 |
| EvalPlus (HumanEval+ & MBPP+) | 53.2 | Pass@1 avg (%) | 50 |
| GSM8K | 37.7 | Accuracy (%) | 43 |
| Big Code Memorization - HumanEval pass@1 | 43.9 | HumanEval pass@1 (%) | 41.2 |
| Big Code Memorization - HumanEval-ET pass@1 | 38.41 | HumanEval-ET pass@1 (%) | 41.2 |
| MMLU | 40.5 | Accuracy (%) | 13.4 |
| ARC Challenge (AI2) | 35.7 | Accuracy (%) | 10.5 |
| WinoGrande | 59.8 | Accuracy (%) | 10 |
Interactive version: theaggregate.ai/model?slug=codeqwen1-5-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.