Qwen3 Coder: benchmark results
Provider: Alibaba. Released 2025-07-23. Access: Open.
Aggregate Rating 1661 ± 17, rank #290 of 1656 rated models, from 40 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FirmCORe - Opportunity Detection | 56.48 | Macro-F1 (%) | 81.2 |
| MT-AgentRisk - Notion | 60 | Attack success rate (%): share of the 15 Notion-MCP tasks th | 70 |
| ProOPF-B (Zero-Shot) - Level 3 | 6.9 | Objective-value accuracy (%) on the 29 Level 3 cases (struct | 68.8 |
| ProOPF-B (Few-Shot) - Level 1 | 80.56 | Objective-value accuracy (%) on the 36 Level 1 cases (explic | 62.5 |
| MT-AgentRisk - Playwright | 67.86 | Attack success rate (%): share of the 140 Playwright-MCP web | 60 |
| MCP-Persona - Rednote-Centric Cross-Server | 23.57 | Checkpoint accuracy (%): mean LLM-judged score (0, 0.5 or 1) | 58.3 |
| FirmCORe - Role Direction | 59.54 | Macro-F1 (%) | 56.2 |
| Quantum API Drift - Qiskit 1.3 | 28.67 | Pass@1 (%; unbiased estimator from 3 samples per task, macro | 53.3 |
| MCP-Persona - Content Management | 13.18 | Checkpoint accuracy (%): mean LLM-judged score (0, 0.5 or 1) | 50 |
| ProOPF-B (Few-Shot) | 23.97 | Objective-value accuracy (%) over all ProOPF-B's 121 expert- | 50 |
| SuperCLUE-WebCoding - Overall | 67.93 | Score | 46.2 |
| EvasionBench | 78.16 | Macro-F1 (%) | 42.3 |
Interactive version: theaggregate.ai/model?slug=qwen3-coder · How It Works · Data refreshed daily, snapshot 2026-10-20.