Release tracker / MiniMax

Compiled by Spring Prompt

MiniMax: latest announcements and model releases

As of 20 November 2026, MiniMax's latest model release is MiniMax-M3.1-Flash, released on 19 Oct 2026 (32 days ago). Before it came MiniMax-M3.1-Flash-Preview, on 28 Sep 2026. We have recorded 8 MiniMax launches since 13 Jun 2025.

Every MiniMax model launch we have recorded, newest first: MiniMax. Each one links to its source, and to how it ranks on our benchmarks where we have tested it.

Latest
MiniMax-M3.1-Flash, 19 Oct 2026
Launches
8
Since
13 Jun 2025
Since last launch
32 days

MiniMax's latest model

MiniMax-M3.1-Flash came out 32 days ago. Spotted a launch we're missing? Tell us.

How often MiniMax ships

Lines with at least three launches in the last two years. Amber: longer since the last launch than the line's average gap. Grey: no launch for 90 days or more, marked quiet.

LineSizeLaunchesAverage gapLatestLast launch
MiniMax Medium 8 70 days MiniMax-M3.1-Flash19 Oct 2026 32 days ago

Every MiniMax release, with its release date

Our reviews of MiniMax models

The launch reviews we have published for MiniMax models, the newest release first.

Questions

What is MiniMax's latest model?
MiniMax-M3.1-Flash, released 19 Oct 2026 (32 days ago). It is the newest of the 8 MiniMax launches we have recorded since 13 Jun 2025.
When will MiniMax release its next model?
MiniMax has not announced its next model. It usually releases a new MiniMax model about every 70 days; the last was MiniMax-M3.1-Flash, 32 days ago.
How often does MiniMax release new models?
Over the last two years: a new MiniMax model about every 70 days (8 launches).

Other providers: OpenAI, Mistral AI, Alibaba, Anthropic, Google, DeepSeek, Meta, xAI, Z.ai, Moonshot AI, Microsoft, NVIDIA, Xiaomi, Amazon, StepFun, Thinking Machines. Or see every launch on one timeline.

A launch is the first day a model was publicly available, previews included; a family's sizes released together count once, and later dated updates of a model are left out. Main lines are checked by hand against the provider's own changelog or announcement where we could read it ("Provider"), otherwise a reference page ("Reference"); the rest come from Epoch AI's model data (CC BY 4.0).