Alibaba's Qwen3.8-Max Reaches 2.4 Trillion Parameters

Alibaba previewed Qwen3.8-Max on 19 July 2026 and launched it on 3 August with published specifications and API pricing. It is the largest and most capable model in the Qwen family to date.

Specifications

A mixture-of-experts model with 2.4 trillion total parameters and 95 billion active per query, built on the Qwen3.5 architecture. The context window is up to 1 million tokens. It is natively multimodal on input - text, images and video - and returns text.

Standing

Alibaba said it would release the model's weights for public download, and made the model widely accessible ahead of that release. Reporting at launch placed it above Moonshot's Kimi K3 on several benchmarks despite Kimi K3's larger 2.8T parameter count.

Roster note: our roster currently carries Qwen3.7 Max. Qwen3.8-Max supersedes it. We will migrate the roster entry once we can verify the full specification set - context tiers, pricing and licence terms - against Alibaba's own documentation rather than launch coverage.