Models·MiniMax

MiniMax M2.5: An Open-Weight Text Model

minimax/minimax-m2.5

MiniMax M2.5 is an open-weight text model with reasoning and tool support.

$0.27Input, per million tokens
$1.08Output, per million tokens
205KContext window
128KLongest answer

205K tokens of context — about 153,600 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Tool use
  • Open weights
  • Takes text
  • Gives text

Released 12 Feb 2026 · checked 3 Oct 2026.

About it

MiniMax M2.5 is a text-only language model from MiniMax, released February 12, 2026. It supports reasoning and tools, uses up to 204,800 tokens of context, and can produce up to 128,000 tokens in one answer. The model is open weight and costs $0.27 per million input tokens and $1.08 per million output tokens.

Review

In a line

MiniMax M2.5 is a text language model made by MiniMax. It was released on February 12, 2026, and its weights are open for use.

What it's good at

The model is aimed at practical digital work and builds on the coding focus of the earlier M2.1. It supports reasoning, which helps it work through multi-step requests.

It also supports tools, so an app can let it take actions or fetch information through connected features. Its input and output are text only. You can compare it with other models with tool calling or open-weight models.

Price in plain terms

Input costs $0.27 per million tokens, and output costs $1.08 per million tokens. Input tokens are the text you send; output tokens are the text it returns.

At the stated rates, reading a 500-word page costs about 0.018 cents. One chat turn with about 1,000 input tokens and 500 output tokens costs about 0.081 cents. A thousand such turns cost about $0.81.

Context and limits

M2.5 can take up to 204,800 tokens in one conversation or request. That is the amount of text it can consider at once.

A single answer can contain up to 128,000 tokens. It handles text input and produces text output, so the facts do not describe image or audio handling.

Ways to use it

People can use M2.5 in apps and chat tools that offer the model.

The model identifier is minimax/minimax-m2.5. Its open weights may suit projects that want to run or adapt an open-weight model.

Verdict

Choose M2.5 if you need a text model with reasoning, tool support, a very large context window, and open weights. Its per-token pricing is low for input and higher for generated text, while the supplied examples keep typical chat costs below one dollar per thousand turns.

Skip it if your work requires images, audio, or another output type, because the listed modality is text only. The available facts also do not specify particular app integrations, so check whether your chosen software offers this model.

Verdict

MiniMax M2.5 fits text-focused work that benefits from long prompts, step-by-step reasoning, and tool connections. Open weights give it a different route from closed models, while the published token rates make everyday chat inexpensive at the example volume. It is not a match for media generation or analysis beyond text. Before choosing it, consider whether your intended software supports the way you want to run an open-weight model.

  • Open weights are available.
  • Supports reasoning and tools.
  • 204,800-token context window.
  • Up to 128,000 output tokens.
  • 1,000 example chat turns cost about $0.81.
  • Text input and output only.
  • Output tokens cost more than input tokens.
  • Pick it ifLong text tasks needing reasoning and tool use; Projects seeking an open-weight text model; High-volume chat at the stated token rates
  • Skip it ifYou need image, audio, or other non-text output.; You require named integrations that are not confirmed here.; You want a model with no open-weight setup considerations.

Pricing

There are no monthly plans in the facts. Usage is billed per million tokens: $0.27 for input and $1.08 for output. The examples work out to about 0.018 cents for one page, 0.081 cents for one chat turn, and $0.81 for 1,000 turns.

Questions

Who makes MiniMax M2.5?

MiniMax makes M2.5. Its lab and model family are also listed as MiniMax and minimax.

When was MiniMax M2.5 released?

It was released on February 12, 2026.

Is MiniMax M2.5 open weight?

Yes. The model is listed as open weight, with the Hugging Face ID MiniMaxAI/MiniMax-M2.5.

Does it support reasoning and tools?

Yes. The facts list both reasoning and tool support.

How much text can it handle at once?

Its context window is 204,800 tokens.

How long can one answer be?

The maximum output is 128,000 tokens.

What does MiniMax M2.5 cost?

Input is $0.27 per million tokens and output is $1.08 per million tokens. A thousand example chat turns cost about $0.81.

What kind of content does it handle?

It accepts text and returns text. No image or audio modality is listed.

On these boards

More from MiniMax