AI Model

MiniMax M2 M2.5 Highspeed

Released on February 12, 2026

Changes from previous version

Same model family as M2.5 with optimized inference for higher TPS; pricing and benchmarks align with the standard M2.5 release.

Release Summary

Higher-throughput M2.5 variant with the same capabilities as MiniMax-M2.5 at faster output token rates for latency-sensitive agent workloads.

Timeline

February 12, 2026

MiniMax M2.5 Highspeed released

MiniMax ships M2.5-highspeed alongside M2.5 on the API platform.