<MCP.Playground/>
online

MiniMax M3.1 Flash

Runs MCP tools

MiniMax’s Flash model from October 2026. Despite the name it is not a cut-down M3: MiniMax and OpenRouter both describe it as an improvement on M3 for agentic tasks, coding and cybersecurity, served about three times faster.

Vendor
MiniMax
Released
19 October 2026
Context
1M tokens
Input
Text, image, video

Paste a server URL, pick a model, and watch it call your tools in a real conversation. No install.

No server of your own? Leave it blank and use one of the public mock servers.

In your browser

MCP Agent Studio

Server has a public URL? Chat with MiniMax M3.1 Flash on it, watch every tool call, and compare models side by side.

Open MCP Agent Studio →

On your computer

MCP Playground Desktop

Server runs on localhost or over stdio? The browser can’t reach it. The desktop app can, with MiniMax M3.1 Flash in Agent Studio. Download free

Mac (Apple Silicon)Mac (Intel)Windowsor Microsoft Store

MCP Evals

Can MiniMax M3.1 Flash actually complete tasks with your server?

One chat shows the tool calls work. An eval writes real tasks from your tool schemas, has MiniMax M3.1 Flash drive each one, and reports where it got the wrong answer even though every call succeeded. Pick MiniMax M3.1 Flash as the driver.

Run evals →See a sample

What MiniMax M3.1 Flash is

MiniMax M3.1 Flash became generally available on 19 October 2026 through the MiniMax API, M Plan and MiniMax Code. MiniMax calls it a next-generation Flash model for agentic tasks, coding, computer use and cybersecurity, with support for chained tool calls over long task sequences. It keeps M3’s 1M-token window and image and video input, with up to 524K tokens of output on OpenRouter. MiniMax says the weights will be released later, without a date in the launch post.

What MiniMax says

  • DeepSWE v1.1: 73.7%.
  • OSWorld V2: 61.7. Agents’ Last Exam: 29.6.
  • BankerToolBench: 76.2, covering multi-step planning, document consistency and financial compliance.
  • CyberGym: 82%.
  • Inference nearly three times faster, above 200 tokens per second in everyday use and above 170 at the median.

These are the vendor’s own claims, not measurements of ours. Run the model against your own server to find out whether they hold for your tools.

How it reaches MCP

  • MCP Playground’s Agent Studio: paste your server URL above and run M3.1 Flash against it in the browser, with no MiniMax key
  • MiniMax’s API, which offers Anthropic-compatible and OpenAI-compatible endpoints, so Claude Code and other MCP clients that accept a custom base URL can use it
  • MiniMax Code, MiniMax’s own coding agent, where M3.1 Flash is available

What to watch for

  • MiniMax published no MCP Atlas score this time, though it did for M3 (74.2%). The tool-heavy benchmarks it did publish, such as BankerToolBench and DeepSWE, are not MCP-specific, so your own server is the real test.
  • MiniMax’s M3 docs say the full assistant message, thinking blocks included, must go back into history in multi-turn tool loops. The M3.1 Flash post doesn’t say whether that changed, so use a client that keeps thinking blocks.
  • OpenRouter lists reasoning_effort for M3.1 Flash, which it does not for M3, along with seed and structured_outputs. A lower effort is worth testing on simple servers where speed matters more than planning.

Run M3 and M3.1 Flash on the same four-step task against the complex-schema mock: create_user_profile, process_order, analyze_data, then configure_workflow. Compare the calls first, then the time each run takes; the speed claim is the main reason to switch if the calls match. All the mock servers →

Available here

ModelModel ID
MiniMax M3.1 Flashminimax/minimax-m3.1-flash

Go deeper

Frequently asked questions

Does MiniMax M3.1 Flash support MCP?
Yes, through any MCP client that can use MiniMax’s Anthropic- or OpenAI-compatible API, such as Claude Code with a base-URL change. You can also run it against your own server in MCP Playground with no MiniMax key.
MiniMax M3.1 Flash or M3?
M3.1 Flash is the newer model. MiniMax describes it as better on agentic tasks, coding and cybersecurity and about three times faster. M3 has the published MCP Atlas score. Both are available here, so run the same prompt on each against your server.
Is MiniMax M3.1 Flash open weight?
Not yet. MiniMax’s launch post says the weights will be released later, without giving a date.
What is the context window of MiniMax M3.1 Flash?
1M tokens, as listed on OpenRouter, the same as M3, with up to 524K tokens of output.

Sources