hybrid attention

MiniMax-M1 brings 1M-token open reasoning model

MiniMax released M1, an open-source reasoning model with 1M-token context, 80k output, and low-cost API pricing.