MiniMax: MiniMax-M2
minimax-m2- Input / Output
- $0.30 / $1.20
- Context
- 200K
- Max output
- 128K
- Modalities
- Chat
- Capabilities
- Tool use
About MiniMax-M2
MiniMax-M2 is MiniMax's open-weight language model built for coding agents and tool-heavy workflows, released in October 2025. MiniMax presents it as an efficient model for the agentic era. It is the first of the M2 line, so later releases (M2.1, M2.5, M2.7) improve on it for programming and productivity. It is text-in, text-out with tool calling and reasoning.
Where it works well
- Built for agentic loops where the model plans, calls tools, reads results and continues.
- Supports reasoning before the answer, which helps on multi-step coding and debugging tasks.
- Open weights, so teams can compare hosted results with a self-hosted deployment of the same model.
- Handles tool calls and prompt caching, which cuts repeated work in long agent sessions.
When to choose another model
- It is the oldest of the M2 models, and M2.1 and later were released specifically to improve coding.
- Text input only, so screenshots or diagrams need a multimodal model such as MiniMax M3.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Text to textResponses APIPOST/v1/responsesAPI format:curl https://api.tokenlab.sh/v1/responses \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-xxx" \ -d '{ "model": "minimax-m2", "input": "Hello!" }'
Pricing
The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.
Default spec
per 1M tokens- Official price
- Input $0.30 / Output $1.20 / Cache read $0.03 / Cache write $0.375
- Verified price
- —
- Discount
- —
Cache read
- Official price
- $0.03
- Verified price
- —
- Discount
- —
Cache write
- Official price
- $0.375
- Verified price
- —
- Discount
- —
| Official priceper 1M tokens | Verified priceper 1M tokens | Discount | |
|---|---|---|---|
| Default spec | Input $0.30 / Output $1.20 / Cache read $0.03 / Cache write $0.375 | — | — |
| Prompt cache pricing | |||
| Cache read | $0.03 | — | — |
| Cache write | $0.375 | — | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open MiniMax-M2 in Console with a prompt ready to edit or send.
Help me try minimax-m2 with a short message at /v1/responses. Show the reply, latency, and cost.
Use cases
Coding agent backend
Drive an editor or terminal agent that reads a repository, edits files and runs commands through tools.
Tool-orchestrating assistants
Let a model decide which internal API to call next across a multi-step support or operations flow.
Baseline for upgrades
Keep it as a reference run when testing whether M2.5 or M2.7 improves your own agent tasks.
Prompt examples
Read the failing CI log below, locate the broken module, and propose a patch with a short explanation.
You can call search_tickets and update_ticket. Close every duplicate of ticket 812 and summarize what you changed.
Refactor this Python function to remove the global state, keeping the public signature unchanged.
This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.
FAQ
What is MiniMax-M2 designed for?
MiniMax describes it as an efficient model for the agentic era. In practice that means coding agents, tool calling and long multi-step tasks, delivered by an open-weight model that also reasons before it answers.
Is MiniMax-M2 open source?
MiniMax released the M2 weights openly, and this line has been published on Hugging Face under the MiniMaxAI organization, so you can read the model card and run the same model yourself.
Does it accept images?
No. It is a text model. Use MiniMax M3, which MiniMax describes as natively multimodal, when images or video must go in; keep M2 for text, code and tool-driven work.
Should I move from M2 to a newer MiniMax model?
For coding work, usually yes. M2.1 targeted programming and refactoring, and M2.5 and M2.7 followed. M2 remains useful as a reference run when you compare results across versions.
Does MiniMax-M2 support tool use and caching?
Yes. Tool use, prompt caching and reasoning are all supported, so it fits agent loops that repeat a large shared prefix, such as a long system prompt or tool list sent on every turn.
How much does MiniMax-M2 cost?
On TokenLab, MiniMax-M2 costs Input $0.30 / Output $1.20 per 1M tokens. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
What are the context window and output limit of MiniMax-M2?
MiniMax-M2 accepts up to 200,000 tokens of context and returns up to 128,000 tokens in one response.
Which endpoint should MiniMax-M2 use?
Use https://api.tokenlab.sh/v1/responses for MiniMax-M2. The request example below shows the matching code shape.
Which operations does MiniMax-M2 support?
MiniMax-M2 supports Text to text. Select an operation above to see its endpoint and request example.
Compare MiniMax-M2
Sources
Reviewed Oct 2, 2026
More from MiniMax