minimax-m2.5
MiniMax M2.5
MiniMax productivity model for complex real-world digital work.
- Input price from
- $0.06/1M tokens
- Output price from
- $0.24/1M tokens
- Context window
- 204800
- Modalities
Pricing
Cache writes: not charged
All prices are per million tokens and already include the discount shown above.
These public prices exclude group discounts. Your billed price may be lower based on your group.
Stability & performance
Success rate
61.85%
Latency
Throughput
56106.29 tok/s
Prompt cache hit rate
10.11%
24-hour status
Based on billed requests over the past 24 hours.
Integration & capabilities
Supported protocols
- Compatible format
- OpenAI
- Routes
POST /v1/chat/completionsGET /v1/models- Features
- StreamingPrompt cachingFunction callingSystem messages
- Request parameters
- modelmessagestemperaturetoolsstream
Connect your tool
Cursor
- BASE URL
- https://api.unbar.sh/v1
- MODEL
- minimax-m2.5
- Create your key on the keys page.
- Open Settings → Models and turn on Override OpenAI Base URL; set it to
https://api.unbar.sh/v1. - Paste that key into the OpenAI API Key field.
- Add a custom model named
minimax-m2.5and select it.
Supported capabilities
The model can call tools you define and return structured arguments to run them.
You can require, allow, or forbid tool use for a single request.
The model can return JSON that matches a schema you supply.
The model can work through a problem step by step before it answers.
Repeated context is cached, so you pay less for tokens the model has already read.
The model can read and describe images sent with a prompt.
The model can read PDF documents sent with a prompt.
The model can drive a computer interface to complete a task.
Context & limits
- Context window
- 204800
- Max output tokens
- 128K