Qwen 3.5 397B A17B Instruct: benchmark results
Provider: Alibaba. Released 2026-02-16.
Aggregate Rating 1721 ± 33, rank #265 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ProactBench - Emergent | 72.9 | Pass rate (%) on the Emergent triggers (turns 1-3: inferring | 100 |
| ProactBench - Weighted Score | 60.6 | Weighted score (%): Pass counts 1, Partial 0.5 and Fail 0, a | 90 |
| Video-MME-v2 - Level 3 Complex Reasoning (w/o subs) | 16.15 | Level 3 Score w/o sub (%) | 85.4 |
| LUNAR - Curated Context | 3.98 | Personalization score (1-5; mean of coverage and depth, GPT- | 83.3 |
| ProactBench - Critical | 63 | Pass rate (%) on the Critical triggers (turns 4-7: synthesis | 80 |
| ProactBench - Recovery | 17 | Pass rate (%) on the Recovery triggers (turns 8-10: grounded | 80 |
| LUNAR - Agentic Memory | 2.83 | Personalization score (1-5; mean of coverage and depth, GPT- | 77.8 |
| LUNAR - Full Context | 3.54 | Personalization score (1-5; mean of coverage and depth, GPT- | 77.8 |
| LUNAR - RAG | 3.05 | Personalization score (1-5; mean of coverage and depth, GPT- | 77.8 |
| Video-MME-v2 | 35 | Avg Accuracy w/o sub (%) | 75 |
| Video-MME-v2 - Level 2 Temporal Understanding (w/o subs) | 18.07 | Level 2 Score w/o sub (%) | 75 |
| Video-MME-v2 - Non-Lin Score (w/o subs) | 16.9 | Non-Lin Score w/o sub (%) | 75 |
Interactive version: theaggregate.ai/model?slug=qwen-3-5-397b-a17b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.