Qwen 3.6 27B (Thinking): benchmark results
Provider: Alibaba. Released 2026-04-22. Access: Open.
Aggregate Rating 1664 ± 12, rank #473 of 2170 rated models, from 137 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MVX-Bench - Counting | 44.7 | Accuracy (%) on the 150 Counting questions (pick the candida | 100 |
| MVX-Bench - Spatial Relation | 29.3 | Accuracy (%) on the 150 Spatial Relation questions (pick the | 100 |
| Medmarks - LongHealth Task 2 | 91.33 | Score (%) | 100 |
| Nejumi 4 - jaster (0-shot) - JSQuAD | 83 | Character F1 (x100) | 99.3 |
| Medmarks - PubMedQA | 81.13 | Score (%) | 98.9 |
| Nejumi 4 - Toxicity - Social Norms | 99.6 | Criteria met (%) | 98.6 |
| Medmarks - MedCalc-Bench | 81.64 | Score (%) | 97.7 |
| Medmarks - SuperGPQA Medicine Easy | 67.77 | Score (%) | 97.7 |
| MVX-Bench - Multi-View Reasoning | 70.8 | Accuracy (%) on the 106 Multi-View Reasoning questions (pick | 97.5 |
| Nejumi 4 - BFCL - Live AST | 75.93 | Accuracy (%) | 97.2 |
| Medmarks - MedHallu Hard | 62.99 | Score (%) | 96.6 |
| Medmarks - MedHallu Medium | 78.62 | Score (%) | 96.6 |
Interactive version: theaggregate.ai/model?slug=qwen-3-6-27b-thinking · How It Works · Data refreshed daily, snapshot 2026-11-05.