LaunchAI ModelsAugust 3, 2026

Alibaba Qwen releases Qwen3.8-Max, a 2.4T-parameter MoE model

Qwen3.8-Max is a 2.4-trillion-parameter mixture-of-experts model accepting text, image, and video input. It matches Kimi K3 and DeepSeek V4 Flash on benchmarks, excelling in coding. Open weights ship next week, along with Qwen3.8-27B.

3 sources

Daily brief

Get tomorrow's AI brief in your inbox

More stories today

Open the live feed