Qwen 3 8B: benchmark results
Can switch between direct answers and explicit reasoning within the same conversation, without loading another checkpoint. Training covers multilingual dialogue, code and tool use. Provider: Alibaba. Released 2025-04-29. Access: Open.
Aggregate Rating 1532 ± 3, rank #674 of 1656 rated models, from 2633 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AWARE-US (Local Weighting) - S2 Two Constraints | 58.5 | Car match (%; share of the 41 persona-grounded car-search di | 100 |
| AWARE-US (Local Weighting) - S3 Any Repair | 47.5 | Car match (%; share of the 40 persona-grounded car-search di | 100 |
| Audio2Tool (Text Oracle) - Correction | 46.8 | Exact match (%): share of queries whose predicted tool or or | 100 |
| Audio2Tool (Text Oracle) - Direct Commands | 85.6 | Exact match (%): share of queries whose predicted tool or or | 100 |
| Audio2Tool (Text Oracle) - Multi-Intent | 38.7 | Exact match (%): share of queries whose predicted tool or or | 100 |
| BAGEL - Xeno-canto Modulation | 38.2 | Accuracy (%, times 100) on the 419 Xeno-canto questions on m | 100 |
| CSSBench - Privacy Leakage | 5.33 | Attack success rate (%; share of obfuscated malicious privac | 100 |
| DynaSchedBench | 1.01 | Mean relative makespan gap (%): makespan above the best sche | 100 |
| Finetuning with Scientific Data Increases Hall | 67.1 | OFS (self-reported) | 100 |
| GeoBenchLLM - GKMC | 82 | Accuracy (%) | 100 |
| GeoBenchLLM - GeoSQA | 63 | Accuracy (%) | 100 |
| GeoBenchLLM - GridRoute | 81 | Optimal route ratio (%) | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-8b · How It Works · Data refreshed daily, snapshot 2026-10-20.