MiniMax M3.1 Flash is a cheaper agent model
MiniMax says the new Flash model is nearly 3× faster and 70% cheaper on agent tasks, at $0.15 per million input tokens. Weights come later.
Official source (opens in a new tab)
MiniMax's 19 October 2026 research post says MiniMax M3.1 Flash is available. MiniMax says an earlier anonymous model, Space Bunny, which processed more than 200 trillion tokens, was an early version of it. The post says online inference is nearly three times faster than the previous Flash model and that agent-task cost is 70% lower. It quotes everyday output speeds of 200+ tokens per second, with a median around 170+. Flat pricing per million tokens is $0.15 input, $0.60 output, and $0.015 cached input.
MiniMax says it scores 73.7% on DeepSWE v1.1 and 61.7 on OSWorld V2, and that it is available through the MiniMax API, M Plan, and MiniMax Code. The post says model weights will be released later.
Who it's for: people calling a fast model for coding and computer-use agents, or trying MiniMax Code.
Fits: Best for Vibe Coding only as a model you might call from an agent. Cursor stays the first pick for an agent that lives in the editor.