Qwen 2.5 Coder 32B Instruct: benchmark results
Alibaba's Apache-2.0 32B code specialist that matched GPT-4o on coding benchmarks - the open-source SOTA coder at release (November 2024). Provider: Alibaba. Released 2024-11-12. Access: Open.
Aggregate Rating 1574 ± 8, rank #528 of 1656 rated models, from 222 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Delulu | 0.85 | pass@1 | 100 |
| Delulu - CodeBLEU | 53.1 | CodeBLEU (0-100): n-gram, syntax-tree and dataflow agreement | 100 |
| Delulu - Edit Similarity | 76.5 | Edit similarity (0-100): character-level normalized Levensht | 100 |
| Delulu - Exact Match | 54.1 | Exact match (%) of the completion with the gold completion, | 100 |
| Open Japanese LLM - CG | 68.47 | Score (%) | 99.8 |
| GSM8K | 93 | Accuracy (%) | 97.8 |
| EvalPlus (HumanEval+ & MBPP+) | 82.1 | Pass@1 avg (%) | 97.6 |
| BigCode Models Leaderboard | 83.2 | HumanEval Python Pass@1 (%) | 96.6 |
| Open LLM Leaderboard - MATH Level 5 | 49.55 | Score | 96.6 |
| BigCodeBench | 49 | Pass@1 (%) | 96 |
| MERA Code - RealCodeJava | 38.59 | pass@1 (%) | 95.4 |
| Open LLM Leaderboard - BBH | 66.25 | Score | 94.4 |
Interactive version: theaggregate.ai/model?slug=qwen-2-5-coder-32b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.