Qwen 3 30B A3B 2507 (Thinking): benchmark results
Qwen 3 30B A3B 2507 thinking snapshot. Provider: Alibaba. Released 2025-07-31. Access: Open.
Aggregate Rating 1564 ± 6, rank #815 of 2168 rated models, from 503 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| QuarkMedBench - Highlight | 49.11 | Highlight score rate (%): share of the Highlight rubric poin | 100 |
| TransClean Generation - Prompt 0 - Any Noise | 3.3 | Noise Rate (%) | 100 |
| TransClean Generation - Prompt 0 - Explanatory Text | 2.81 | Explanatory Noise Rate (%) | 100 |
| TransClean Generation - Prompt 0 - Wrong Language | 0.65 | Wrong-language Rate (%) | 100 |
| TransClean Generation - Prompt 1 - Any Noise | 3.26 | Noise Rate (%) | 100 |
| TransClean Generation - Prompt 1 - Wrong Language | 0.51 | Wrong-language Rate (%) | 100 |
| MERA Industrial - ruTXTMedQFundamental - General chemistry | 78.9 | Score (%) | 99.5 |
| MERA - MathLogicQA | 99.74 | Accuracy (%) | 98.6 |
| MERA - ruMultiAr | 100 | EM (%) | 98.6 |
| MERA Code - ruHumanEval | 79.88 | pass@1 (%) | 98.5 |
| MERA Industrial - ruTXTMedQFundamental - Microbiology | 91.1 | Score (%) | 98.5 |
| MERA Industrial - ruTXTMedQFundamental - Pharmacology | 86.7 | Score (%) | 97.6 |
Interactive version: theaggregate.ai/model?slug=qwen-3-30b-a3b-2507-thinking · How It Works · Data refreshed daily, snapshot 2026-10-20.