MiniMaxLaunchAI ModelsOctober 23, 2026

MiniMax releases open weights for 428B-parameter M3.1 Flash

Read original source →technode.com

The mixture-of-experts model activates 23B parameters per token and takes text, image and video inputs with a context window up to 1 million tokens. MiniMax reports 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2 and 76.2 on BankerToolBench.

4 sources

More stories today

Open the live feed