How to Use the MiniMax M3.1 Flash API

Disclosure: This article was written with AI assistance and reviewed by EmpirioLabs AI.

MiniMax M3.1 Flash is available on EmpirioLabs. It is MiniMax's new Flash model for coding and agentic work, released on October 19, 2026. It reads text, images, and video in a 1M token context, writes up to 524,288 output tokens, always reasons before it answers at a depth you choose, and can call your functions and search the web. MiniMax reports stronger agentic, coding, and computer-use results than its earlier models.

Watch the launch film

Every program in the film was written by MiniMax M3.1 Flash in one shot and is shown running as written. The score was generated by ACE-Step on EmpirioLabs. Watch it on YouTube.

What the model does

Text, image, and video input. Send screenshots, diagrams, photos, and video clips in the same request as the prompt, using the standard image_url and video_url content blocks.

Adjustable reasoning depth. The model always reasons. reasoning_effort sets how deeply: low, medium, high, xhigh, or max, with max as the default. The reasoning text comes back in reasoning_content.

Tools and web search. Function calling works on its own or together with the built-in web search, which you turn on with tool_web_search. Repeated prompt prefixes are cached automatically and billed at the cached input rate.

How to call MiniMax M3.1 Flash

Use the OpenAI-compatible chat completions endpoint and set model to minimax-m3-1-flash. This request searches the web and reasons at medium depth:

curl https://api.empiriolabs.ai/v1/chat/completions \
  -H "Authorization: Bearer $EMPIRIOLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-m3-1-flash",
    "reasoning_effort": "medium",
    "tool_web_search": true,
    "messages": [
      {"role": "user", "content": "What changed in the latest stable Python release? Three bullet points."}
    ]
  }'

The answer ends with a Sources list of the pages the model used, and usage.tool_usage.web_search counts the searches it ran. The same model ID works on the Responses, Messages, and Gemini-compatible endpoints listed on the model page, and with "stream": true.

Choosing a reasoning effort

Depth and output tokens. Higher levels reason longer before answering. On the same set of problems in our tests, max used about twice the reasoning tokens of low, and reasoning tokens are billed as output tokens. Use low or medium for quick replies and routine tool calls, and keep max for hard coding and multi-step work.

Moving from MiniMax M2.7

Retirement date. MiniMax M2.7 and MiniMax M2.7 Highspeed retire on November 17, 2026. MiniMax M3.1 Flash is their successor: change the model ID to minimax-m3-1-flash. It also adds image and video input and a context window five times larger.

Operational notes

• Reasoning cannot be turned off. enable_thinking has no effect, and a reasoning_effort outside the five levels is ignored, so the request reasons at max. Set low for the shortest reasoning.
• Search results count as input tokens. The pages a search returns are added to the model's input, so a searched request can bill well over 100,000 input tokens, and one request can run several searches.
• Structured output is not supported. response_format is ignored. Ask for JSON in the prompt when you need it.
• Search answers end with a Sources list. Set disable_formatting to true when another program reads the output and expects the answer alone.

Pricing

Token usage is billed at one input rate, one output rate, and a lower rate for cached input, the same at every context length. Reasoning tokens are billed as output tokens, and each executed web search adds a per-search fee. See the live model page and pricing page for current rates.

Start building

Try MiniMax M3.1 Flash in the Playground, read the API documentation, or compare it with other models in the model catalog.