Qwen 3 Coder 30B A3B Instruct: benchmark results
Alibaba's open 30B A3B MoE coder (3.3.3B active) tuned for agentic coding, with 256K native context extendable to 1M (July 2025). Provider: Alibaba. Released 2025-07-23. Access: Open.
Aggregate Rating 1607 ± 10, rank #415 of 1656 rated models, from 147 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| TestExplora (Black Box) - Entry Coverage | 52.98 | Entry coverage (%): line coverage of the functions under tes | 100 |
| AutoEval-ECov (Agentic) | 83.8 | Coverage pass rate (%): share of the 156 problems of AutoEva | 84.6 |
| MCPGen - Tool Creation (Individual Pass) | 58.5 | Individual pass rate (%; share of the 911 withheld workflow | 80 |
| AutoEval-ECov (Agentic) - Average Coverage | 94 | Average coverage (%) of the model's SystemVerilog testbenche | 76.9 |
| AutoEval-ECov (Direct Inference) | 75.3 | Coverage pass rate (%): share of the 156 problems of AutoEva | 76.9 |
| AutoEval-ECov (Direct Inference) - Average Coverage | 87.1 | Average coverage (%) of the model's SystemVerilog testbenche | 76.9 |
| CVDP-ECov (Agentic) | 63.9 | Coverage pass rate (%): share of the 83 hardware repositorie | 76.9 |
| BIRD-Python - SQL | 61.6 | LLM-validated execution accuracy (%; generated SQL executed | 75 |
| MERA Code - RealCodeJava | 30.87 | pass@1 (%) | 74.6 |
| Tinybird AI SQL Benchmark - Success Rate | 100 | Questions answered with a valid query within 3 retries (%) | 71.7 |
| PAIR-Bench | 90.9 | Final fix rate (%; share of the 440 wrong-answer Codeforces | 71.4 |
| PAIR-Bench - Gap Closure | 94.25 | Gap closure (%; share of the remaining hidden-test pass-rate | 71.4 |
Interactive version: theaggregate.ai/model?slug=qwen-3-coder-30b-a3b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.