MiniMax M3.1 Flash
MiniMax's fast agent model, released October 19, 2026. Flat pricing of $0.15 in and $0.60 out per million tokens, with output speeds above 200 tokens a second.
- Made by
- MiniMax
- Released
- Type
- Fast & cheap
- Status
- Live
- How to get it
- MiniMax API, M Plan and MiniMax Code; MiniMax says the weights will follow
- Good for
- Cheap agentsCodingComputer use
What it is
M3.1 Flash is MiniMax’s model for agents, coding and computer use, rolled out on October 19, 2026. Before launch, an early version ran anonymously under the name “Space Bunny” and processed more than 200 trillion tokens. MiniMax says it is close to three times faster than its earlier models, with typical output speeds of 170 to 200+ tokens a second.
What it costs
$0.15 per million input tokens and $0.60 per million output tokens, with cached input at $0.015. That makes it one of the cheapest capable models on this list. MiniMax is moving its older M2.5 and M2.7 plan users onto it automatically.
What it means for you
If speed and cost matter more than squeezing out the last bit of quality, for example a support bot or an agent that runs all day, M3.1 Flash is worth testing against bigger models on your own task.
Use cases this model is good for
How to automate a boring computer task with an AI-written script (no coding experience needed)
You do the same file-renaming or copy-paste job every week and suspect a small script could do it for you.
Best pick: ClaudeHours saved
