Qwen 3 32B: 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 1601 ± 3, rank #431 of 1635 rated models, from 2325 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AMI-ME - Meeting Effectiveness Scoring | 0.64 | Segment-level Spearman correlation (-1 to 1) between the mod | 100 |
| AMI-ME - Meeting Effectiveness Scoring (Non-Scenario Meetings) | 0.46 | Segment-level Spearman correlation (-1 to 1) between the mod | 100 |
| AMI-ME - Meeting Effectiveness Scoring (Scenario Meetings) | 0.66 | Segment-level Spearman correlation (-1 to 1) between the mod | 100 |
| AstroMind - Intent Inference | 67.74 | Accuracy (%; 66 maneuver-intent questions, 31 four-option an | 100 |
| CreativityBench | 25.88 | Gold correct rate (%, times 100): share of tasks where the m | 100 |
| CreativityBench - Entity Correct | 62.46 | Entity correct rate (%, times 100): share of tasks where the | 100 |
| DRNoise - Conditional Deference | 61.3 | Conditional deference (%; share of the tasks a model answere | 100 |
| ESF-Bench - Multi-Source | 87.3 | Tagged slot accuracy (%; slots tagged with a MS (Multi-Sourc | 100 |
| ESF-Bench - Personalization | 66.8 | Tagged slot accuracy (%; slots tagged with a PER (Personaliz | 100 |
| ESF-Bench - Relative | 91.2 | Tagged slot accuracy (%; slots tagged with a REL (Relative) | 100 |
| FDARxBench - Multi-Hop | 47.3 | Answer accuracy (x 100) on cross-section multi-hop questions | 100 |
| FHIR-AgentBench (Single-Turn + Retriever) | 23 | Answer correctness (%; o4-mini judge of the free-text answer | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-32b · How It Works · Data refreshed daily, snapshot 2026-10-15.