Qwen 2.5 Coder 7B: benchmark results
Alibaba's 7B code-specialized base model in the Qwen2.5-Coder series, pretrained on 5.5 trillion tokens of code and text with a 128K-token context. Provider: Alibaba. Released 2024-09-19. Access: Open.
Aggregate Rating 1381 ± 15, rank #1286 of 1656 rated models, from 43 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| DomainCodeBench | 89.77 | Composite Score (%) | 100 |
| GSM8K | 83.9 | Accuracy (%) | 81.7 |
| Open Korean LLM - KorNAT-Helpful | 49.28 | Normalized accuracy (%) | 78.1 |
| DuckDB-NSQL | 58.7 | Execution Accuracy (%) | 69.8 |
| Open LLM Leaderboard - MATH Level 5 | 19.18 | Score | 69.5 |
| Open LLM Leaderboard - MMLU-Pro | 36.79 | Score | 58.8 |
| ARC Challenge (AI2) | 60.9 | Accuracy (%) | 56.6 |
| Open Korean LLM - Ko-IFEval | 29.59 | Strict accuracy, prompt/instruction mean (%) | 52.3 |
| HellaSwag | 76.8 | Accuracy (%) | 48.7 |
| Open Korean LLM - Ko-GPQA-Diamond | 23.23 | Normalized accuracy (%) | 48.1 |
| GSMA Open-Telco - TeleTables | 29.4 | Score (%) | 47.8 |
| MMLU | 68 | Accuracy (%) | 47.8 |
Interactive version: theaggregate.ai/model?slug=qwen-2-5-coder-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.