Qwen 3.6 35B A3B (Non-reasoning): benchmark results
Qwen 3.6 35B A3B evaluated with reasoning disabled. Provider: Alibaba. Released 2026-04-16. Access: Open.
Aggregate Rating 1554 ± 14, rank #856 of 2168 rated models, from 116 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Werewolf Accusation Benchmark - Role Separation | 1.5 | Belief separation (wolves minus villagers, -3..+3 scale) | 84.6 |
| Trip+ - User Simulation | 51.41 | Simulated traveler experience (0-100), a rate scaled by 100: | 82.4 |
| EPIC-Bench - Embodied Compositional Attributes | 40.01 | Target localization score (0-100) on the embodied compositio | 80.7 |
| AA TAU-2 Bench | 85.09 | Accuracy (%) | 77.4 |
| Tau2-Bench Telecom | 85.09 | Success Rate (%) | 77.4 |
| TableVista - Screen Capture | 46.1 | Accuracy (%) on the 3,000 TableVista questions under the vis | 75 |
| AA Omniscience - Software Engineering (SWE) - Rust | 56 | Accuracy (%) | 74.1 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 28.89 | Accuracy (%) | 73.4 |
| EPIC-Bench - Spatial-Related Attributes | 46.19 | Target localization score (0-100) on the spatial-attribute t | 72.7 |
| TableVista - Simulated Photo | 44.5 | Accuracy (%) on the 3,000 TableVista questions under the vis | 71.4 |
| CDR-Bench - Atomic Filters | 78.05 | Recipe success RS@3 (%) on the atomic filter (single KEEP/DR | 70 |
| EPIC-Bench - Basic Attributes | 46.4 | Target localization score (0-100) on the basic-attribute tas | 69.9 |
Interactive version: theaggregate.ai/model?slug=qwen-3-6-35b-a3b-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.