Qwen3.8 Max is Alibaba's most capable model to date, a 2.4-trillion-parameter sparse Mixture-of-Experts transformer with roughly 95 billion parameters active per token. Announced at the World AI Conference in Shanghai on July 19, 2026, and launched August 3, it processes text, images, video, and documents within a 1M-token context window and outputs up to 131K tokens. Priced at $2/$6 per million tokens, it targets the frontier tier while undercutting competitors. Notably, Alibaba has pledged to release the weights on Hugging Face — the first time a Max-class Qwen model goes open.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| SWE-Bench | 85.6 | accuracy | 2mo ago |
| IFBench | 82.8 | accuracy | 2mo ago |
| LiveCodeBench | 87.9 | accuracy | 2mo ago |
| GPQA Diamond | 92.6 | accuracy | 2mo ago |
| MMLU-Pro | 88.6 | accuracy | 2mo ago |
| Humanity's Last Exam | 43.6 | accuracy | 2mo ago |