MiniMax
Operational· checked every few minutes

MiniMax M2.5 API

The MiniMax M2.5 API for low-cost coding agents, 68% below MiniMax's list price

Price per 1M tokens68%off
Input
$0.10$0.30
Output
$0.39$1.20
Cached input
$0.01
Get your API keyTry it in the playground
Model IDminimax-m2.5
chat.py
from openai import OpenAI
client = OpenAI(
    base_url="https://api.zurelay.com/v1",
    api_key="YOUR_ZURELAY_KEY",
)
stream = client.chat.completions.create(
    model="minimax-m2.5",
    messages=[{"role": "user", "content": "Explain quicksort in two sentences."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="")

Base URL https://api.zurelay.com/v1. Works with any OpenAI SDK. Chat API docs

Context window
205K tokens
Max output
128K tokens
Reasoning
Always thinks
Input
Text
Released
Feb 12, 2026
Status
Operational

Pricing

MiniMax M2.5 API pricing. A fraction of the list price.

Pay as you go from prepaid credit, with no subscription. Top up from $10. Requests that fail are never billed.

RateZurelayMiniMax list
Input
per 1M tokens
$0.10$0.30
Output
per 1M tokens
$0.39$1.20
Cached input
per 1M tokens
$0.01—

One rate at every prompt length, up to the full 205K context window. Streaming, tool calls and structured outputs cost nothing extra.

Every MiniMax model and price
Savings calculator

Move the sliders to your monthly usage.

Input tokens per month50M
Output tokens per month10M
Zurelay
$8.90
MiniMax list price
$27.00
You save every month$18.10

$217.20 a year

Overview

What is MiniMax M2.5?

MiniMax M2.5 is MiniMax's open-weight model for coding, agentic tool use and search, and office work, released on February 12, 2026. The Zurelay MiniMax M2.5 API serves it through one OpenAI-compatible endpoint with a 204,800-token context window and up to 131,072 output tokens per answer, at 68% below MiniMax's list price.

MiniMax released M2.5 on February 12, 2026. It builds on the coding strengths of MiniMax M2.1 and extends them to office work in Word, PowerPoint and Excel, including financial models. MiniMax published the weights on Hugging Face: a Mixture-of-Experts model of about 229B parameters. On Zurelay the MiniMax M2.5 API model ID is minimax-m2.5, and it costs $0.10 per 1M input tokens and $0.39 per 1M output tokens.

MiniMax's reported results center on coding and agents. M2.5 scores 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench, 55.4% on SWE-Bench Pro and 76.3% on BrowseComp with context management. MiniMax says it finishes SWE-Bench Verified 37% faster than M2.1, matching Claude Opus 4.6's speed, and needs about 20% fewer search rounds than M2.1 on its search tests. In office work it won 59.0% of comparisons against other mainstream models on GDPval-MM, MiniMax's internal evaluation. These are MiniMax's own numbers, not independent ones.

MiniMax trained M2.5 with reinforcement learning in hundreds of thousands of real-world environments, across more than 10 programming languages, on Forge, its in-house framework for training agents. It reports that M2.5 learned to plan like a software architect, laying out features, structure and UI design before it writes code, and that it covers full-stack projects for Web, Android, iOS and Windows.

M2.5 always reasons before it answers, and on Zurelay the reasoning comes back in the reasoning field, beside the answer. reasoning_effort is accepted, but every level runs the same way, and none can't switch the reasoning off. It reads and writes text, holds 204,800 tokens of context and writes up to 131,072 tokens per answer. It calls tools, follows a JSON schema in response_format and streams; in our tests it wrote about 100 tokens a second. You pay $0.10 per 1M input tokens and $0.39 per 1M output tokens against MiniMax's list price of $0.30 and $1.20, with cached input at $0.01.

Strengths

Where MiniMax M2.5 shines. And what teams build with it.

01

Coding scores at a low price

MiniMax reports 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench and 55.4% on SWE-Bench Pro.

02

Fast answers

About 100 tokens a second in our tests, and MiniMax reports it finishes SWE-Bench Verified 37% faster than M2.1.

03

Office work

Built with professionals in finance, law and the social sciences for Word, PowerPoint and Excel tasks, including financial models.

04

Long answers

Up to 131,072 output tokens per answer in a 204,800-token window, for long patches, reports and files.

Use cases

  • Coding agents

    Multi-file fixes, refactors and terminal tasks in agent harnesses, at a low price per token.

  • Search agents

    Browse and call search tools over many steps. MiniMax reports 76.3% on BrowseComp with context management.

  • Office documents

    Write reports, slide content and spreadsheet models from your notes and data, with your own tools creating the files.

  • High-volume text work

    Extraction, classification and summaries at scale, with JSON schema output for the next step.

Get started

Call MiniMax M2.5 in three steps.

No waitlist and no new SDK. If your code already talks to OpenAI, it already talks to Zurelay.

  1. 1

    Create a key

    Sign up, add credit and create an API key. Set a monthly budget or a rate limit per key if you like.

    Get your API key
  2. 2

    Point your SDK at Zurelay

    Change the base URL. Everything else in your code stays the same.

    https://api.zurelay.com/v1
  3. 3

    Use minimax-m2.5

    Send chat completions as usual. Streaming, tool calls and usage reporting work as you expect.

    minimax-m2.5

Works with the tools you already use

Compare

Similar models. All on the same key.

57%off

MiniMax M3

$0.13 / $0.51 per 1M tokens

Pick MiniMax M3, MiniMax's newer model, for a 1M-token context, image input and higher coding scores: 59.0% on SWE-Bench Pro against M2.5's 55.4%.

MiniMax M3 API
60%off

MiniMax M3.1 Flash

$0.06 / $0.24 per 1M tokens

Pick MiniMax M3.1 Flash, MiniMax's newest model, for a 1M-token context and image and video input at a lower price per token.

MiniMax M3.1 Flash API
73%off

GLM-5.3

$0.38 / $1.21 per 1M tokens

Pick GLM-5.3, Z.ai's flagship, for harder text-only coding work with a 1M-token context and a choice of reasoning effort.

GLM-5.3 API

FAQ

MiniMax M2.5 API questions.

Anything else? Ask our team and we’ll answer by email.

How much does the MiniMax M2.5 API cost?

On Zurelay, minimax-m2.5 costs $0.10 per 1M input tokens and $0.39 per 1M output tokens, with cached input at $0.01. MiniMax's list price is $0.30 input and $1.20 output per 1M, so you save 68%. Reasoning tokens count as output.

Is there a cheaper MiniMax M2.5 API?

Yes. Zurelay serves minimax-m2.5 at 68% below MiniMax's list price, at one rate for every prompt length. You pay from prepaid credit with no subscription, and failed requests are never billed.

How do I call the MiniMax M2.5 API with the OpenAI SDK?

Install the official openai package, set base_url to https://api.zurelay.com/v1 and use your Zurelay API key. Then send a Chat Completions request with model set to "minimax-m2.5". Zurelay also accepts MiniMax's own name, MiniMax-M2.5, and minimax/minimax-m2.5, so existing code only needs a new base URL and key. MiniMax recommends temperature 1.0 and top_p 0.95.

What is the MiniMax M2.5 context window?

204,800 tokens, and one answer can run to 131,072 output tokens. Reasoning and the answer share that output budget, so leave headroom on hard prompts. In our tests it worked through a 156K-token prompt. M2.5 takes text as input and writes text.

Can MiniMax M2.5 read images or files?

No. On Zurelay M2.5 takes text only: images, PDFs, audio and video aren't supported, and a request that includes one is turned away with a clear error before anything is billed. For images, use MiniMax M3; for images and video, use MiniMax M3.1 Flash.

Can I turn off MiniMax M2.5's reasoning?

No. M2.5 always reasons before it answers, and the reasoning comes back in the reasoning field, beside the answer. reasoning_effort is accepted, but every level runs the same way, and none doesn't switch the reasoning off. For answers without thinking, use MiniMax M3, which switches it off with reasoning_effort none.

Does MiniMax M2.5 support tool calling, JSON and streaming?

Yes. Pass tools as you would to any OpenAI model; a tool forced by name is called, and tool_choice "none" works. Send response_format with json_object or json_schema and the answer comes back as JSON. Set stream: true, and add stream_options.include_usage to get token counts in the final chunk.

How is MiniMax M2.5 different from MiniMax M3?

M3 is MiniMax's newer model. It reads images as well as text, holds 1,048,576 tokens of context against 204,800 for M2.5, and scores higher on MiniMax's coding tests: 59.0% against 55.4% on SWE-Bench Pro, and 80.5% against 80.2% on SWE-Bench Verified. M2.5 takes text only and costs less per token.

Is this the real MiniMax M2.5?

Yes. Requests for minimax-m2.5 run on MiniMax's open-weight M2.5, and Zurelay doesn't quietly swap in a different model: if you turn on smart routing and M2.5 is down, the response names the model that answered. Responses are sampled, so wording varies between runs. Zurelay is independent and is not affiliated with or endorsed by MiniMax.

More models on Zurelay

All models

Stop paying list price.
Start saving today.

Create a key in seconds, point the SDK you already use at Zurelay, and every request costs up to 90% less from the first token.

Get your API keyTalk to an engineer