Qwen 3.8 Max (High): benchmark results
Provider: Alibaba. Released 2026-08-03. Access: API.
Aggregate Rating 1796 ± 21, rank #146 of 2170 rated models, from 60 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GameLogicBench - Atom | 61.9 | Solve rate (%; the 21 Atom tasks (one mechanic in a minimal | 95.8 |
| FrontierCS 2 - Differentiable Population Statistics | 5.67 | Task score (0–100) | 94.1 |
| BVB | 66.24 | Overall (%) | 94 |
| GameLogicBench | 44.44 | Solve rate (%; all 72 tasks; one sample per model-scaffold c | 91.7 |
| GameLogicBench - Combo | 50 | Solve rate (%; the 28 Combo tasks (several interacting mecha | 91.7 |
| FrontierCS 2 - Future Video Sampling | 72.42 | Task score (0–100) | 88.2 |
| FrontierCS 2 - GPU vLLM Serving Optimization | 22.26 | Task score (0–100) | 88.2 |
| FrontierCS 2 - Persistent Treap Update Engine | 34.22 | Task score (0–100) | 88.2 |
| FrontierCS 2 - Robust Image Tokenizer | 57.96 | Task score (0–100) | 88.2 |
| FrontierCS 2 - GPU DBSCAN Kernel Optimization | 69.49 | Task score (0–100) | 82.4 |
| FrontierCS 2 - Approximate-MIPS Query Stopping | 84.93 | Task score (0–100) | 76.5 |
| FrontierCS 2 - Progent Privilege Control | 50.44 | Task score (0–100) | 70.6 |
Interactive version: theaggregate.ai/model?slug=qwen-3-8-max-high · How It Works · Data refreshed daily, snapshot 2026-11-05.