Qwen 3.6 Max: benchmark results
Alibaba's closed-weights Qwen 3.6 Max Preview flagship with a 256K context window. Provider: Alibaba. Released 2026-04-20. Access: API.
Aggregate Rating 1766 ± 14, rank #96 of 1656 rated models, from 54 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FORTIS - Skill-Grounded Tool Selection Fail Rate | 45.2 | Fail rate (%), the share of over-privileged or no-action out | 100 |
| THOR Finding Triage - False Review Load | 22.7 | False positives not suppressed (%, lower is better) | 100 |
| PerfCodeBench | 66.32 | CGRE (self-reported) | 94.7 |
| FirmCORe - Opportunity Detection | 70.23 | Macro-F1 (%) | 93.8 |
| FORTIS | 51.3 | Task 1 EM (self-reported) | 88.9 |
| FORTIS - Skill Selection Fail Rate | 38.5 | Fail rate (%), the share of over-privileged or no-action out | 88.9 |
| FORTIS - Skill-Grounded Tool Selection | 25.9 | Exact-match rate (%) with the minimum-privilege tool set on | 88.9 |
| THOR Finding Triage - CW% | 65.4 | Confidence-weighted classification score (%) | 88.2 |
| GroupTravelBench | 9.66 | GU (Group Utility) (self-reported) | 87.5 |
| LiveSQLBench | 33.79 | Success Rate (imported) | 84.4 |
| PerfCodeBench - Efficiency Gap Closed | 63.1 | Share of comparable tasks (%) whose correct code closes at l | 84.2 |
| CDR-Bench - Order-Sensitive Filter Consistency | 8.27 | Group-level order consistency OCS@3 (%) on the filter-order- | 80 |
Interactive version: theaggregate.ai/model?slug=qwen-3-6-max · How It Works · Data refreshed daily, snapshot 2026-10-20.