Gemini 4 Argon
Analyze multimodal inputs and generate text for complex coding, scientific reasoning, and tool-connected agent applications.
MiniMax M3.1 Flash
www.minimax.io
Downloadable weights support self-hosted multimodal coding and agent workflows, with deployment guidance and up to one million tokens of context under the MiniMax Community license.
Visit website
Pricing
Paid
Available on
API, MiniMax Code, M Plan
Category
Llm Models
What is MiniMax M3.1 Flash?
MiniMax M3.1 Flash is a model for software development, computer operation and multi-step automation. It is aimed at developers building assistants that work with code, interact with applications and coordinate tool calls across longer tasks.
The model combines a 1,000,000-token context window with adjustable thinking depth. Access is available through the MiniMax API, M Plan and MiniMax Code, giving developers a choice between integrating the model into an application and using MiniMax's coding client.
Key features
Coding and agent workflows
Supports software development and multi-step workflows that combine planning, computer interaction and tool calls.
Large context window
A 1,000,000-token context window accommodates extensive task material. This capacity describes context rather than a maximum answer length.
Multimodal API inputs
The Anthropic-compatible interface accepts text, images, video, tool calls and tool results.
Adjustable thinking effort
Choose low, medium, high, xhigh or max effort to control thinking depth. Max is the default.
MiniMax Code access
MiniMax Code provides macOS and Windows desktop clients and a CLI. Its iOS and Android access supports continuing desktop workflows.
Who is it for?
Software feature development
Developers can assign a scoped feature change that involves reading existing code, proposing edits and checking the result.
Application workflow automation
Operations teams can build an agent to coordinate a repeatable task across connected business applications.
Visual task context
Assistant builders can supply screenshots or video alongside instructions when a task depends on visual information.
How to use MiniMax M3.1 Flash
- Register or sign in to the MiniMax API Platform.
- Choose pay-as-you-go access or M Plan, obtain the corresponding key and ensure billing resources are available.
- Configure a compatible SDK with MiniMax's endpoint and select MiniMax-M3.1-Flash.
- Send a scoped task with relevant inputs, choose an effort level and preserve complete assistant responses in tool conversations.
MiniMax M3.1 Flash pricing
Paid
The API uses token-based, pay-as-you-go billing deducted from account balance. M Plan provides a separate subscription route, with consumption depending on the model, context length and task complexity.
M3.1 Flash API
USD 0.15 per million input tokens; USD 0.60 per million output tokens; USD 0.015 per million cached-input read tokens
Usage-based; deducted from account balance as consumption occurs
Token consumption rates for MiniMax-M3.1-Flash API access, separate from M Plan subscription billing.
What to know before you use it
- Thinking cannot be disabled, and thinking tokens count toward the max_tokens allowance.
- Anthropic API compatibility is partial, and some parameters are ignored. Tool conversations must retain complete assistant responses, including thinking and tool-use blocks.
- Downloadable model weights are forthcoming. Current access routes do not provide downloadable weights for local deployment.
Frequently asked questions
Which API model name should I use?
Use MiniMax-M3.1-Flash. The retired MiniMax-M3.1-Flash-Preview name now routes to the released model.
Are M Plan keys interchangeable with pay-as-you-go keys?
They access different billing resources. Use a pay-as-you-go API key for account-balance usage or the subscription key associated with M Plan resources.
Does M Plan include M3.1 Flash?
Yes. MiniMax lists M3.1 Flash in the Go, Explore and Build plans.