MiniMax

(Shanghai, China)

www.minimaxi.com

Creators of MiniMax M3 and the next-generation M3.1 Flash (agentic, computer-use and cybersecurity gains at 70% lower task cost) - a favorite for autonomous agents and batch processing.

2 MODELS TRACKEDNEWEST: MINIMAX M3.1 FLASH · 5D AGO

What is the latest AI model from MiniMax?

MiniMax's primary flagship model is MiniMax M3 (MiniMax MoE (MSA Architecture), 1,048,576 tokens context). Top-5 OpenRouter volume; the default for OpenClaw-style autonomous agents, balancing reasoning depth with batch throughput. Their latest release is MiniMax M3.1 Flash (Agent Flash, released 2026-10-19). How we define “latest”

Release Timeline

NEWEST FIRST

October 2026[1 release]

MiniMax M3.1 Flash

FLASH AGENT PICKOct 19, 2026

MiniMax's next-generation Flash (Oct 19, 2026): major agentic leaps - DeepSWE v1.1 73.7%, OSWorld V2 61.7 and ALE 29.6 (first among Chinese models), CyberGym 82% - at $0.15/$0.60 per 1M tokens (flat pricing, 70% lower task cost than M3) with 200+ TPS output; served anonymously as 'Space Bunny' through 200T tokens; weights to follow.

Agent FlashOPEN DATASHEET

May 2026[1 release]

MiniMax M3

AGENT FAVORITE• FlagshipMay 31, 2026

Top-5 OpenRouter volume; the default for OpenClaw-style autonomous agents, balancing reasoning depth with batch throughput.

Agentic FrontierOPEN DATASHEET