ModelBeat
Models/MiniMax-M2.5
All models
M
minimax/minimax-m2.5

MiniMax-M2.5 is MiniMax’s open-weights AI model, released Feb 12, 2026. It costs $0.30/M input and $1.20/M output tokens, and handles a 205K-token context window.

MiniMax-M2.5 is a SOTA large language model designed for real-world productivity.

Context
205K
Input
$0.30 / 1M
Output
$1.20 / 1M
Cheapest via Venice
$0.27 / $0.95 / 1M
Released
Feb 12, 2026
Modalities
texttext

Benchmarks

Scores on standardized evaluations. Higher is better — and rank shows where MiniMax-M2.5 lands among all models tracked on Model Beat.

33.0
Intelligence Index
Epoch AI
33rd percentile of tracked models
35.0
Coding Index
Epoch AI
35th percentile of tracked models
41.0
Agentic Index
Epoch AI
41st percentile of tracked models

Reasoning

4 evals
ARC-AGI63.7%

Abstract visual reasoning

ARC-AGI-24.9%

Harder abstract reasoning

GPQA Diamond84.8%

Graduate-level scientific reasoning

Humanity's Last Exam20.5%

Frontier of human expert knowledge

Coding

2 evals
SciCode42.6%

Scientific research coding

WebDev Arena1384

Human-rated web development

Agentic & Tools

3 evals
APEX6.2%

Multi-step agentic tasks

Terminal-Bench42.7%

Command-line agentic tasks

τ²-bench95.3%

Tool-agent-user reliability

Changelog

6
  • MiniMax-M2.5: cheapest credible provider via Venice now $0.27 per 1M input tokens (was $0.22, +23%)cheapest provider · Aug 21, 2026
  • MiniMax-M2.5: cheapest credible provider via Venice now $0.95 per 1M output tokens (was $0.9, +6%)cheapest provider · Aug 21, 2026
  • MiniMax-M2.5: vendor list price cut from $0.6 to $0.3 per 1M input tokens (-50%)list price · Aug 15, 2026
  • MiniMax-M2.5: vendor list price cut from $2.4 to $1.2 per 1M output tokens (-50%)list price · Aug 15, 2026
  • MiniMax-M2.5: cheapest credible provider via Inceptron now $0.22 per 1M input tokens (was $0.15, +47%)cheapest provider · Aug 8, 2026
  • MiniMax-M2.5: Humanity's Last Exam improved from 19.1% to 20.5% (+7%)benchmark · Aug 6, 2026

Compare MiniMax-M2.5 with

Frequently asked questions

What is MiniMax-M2.5?

MiniMax-M2.5 is a SOTA large language model designed for real-world productivity.

Who created MiniMax-M2.5?

MiniMax-M2.5 is an AI model developed by MiniMax, part of the MiniMax family, released Feb 12, 2026.

How much does MiniMax-M2.5 cost?

MiniMax-M2.5's list price from MiniMax is $0.30 per million input tokens and $1.20 per million output tokens. The cheapest credible third-party provider on OpenRouter (Venice) serves it at $0.27/$0.95 per 1M.

What is MiniMax-M2.5's context window?

MiniMax-M2.5 supports a context window of up to 205K tokens.

How does MiniMax-M2.5 perform on benchmarks?

On standardized evaluations tracked by Epoch AI, MiniMax-M2.5 scores 95.3% on τ²-bench, 42.6% on SciCode, 84.8% on GPQA Diamond.

Is MiniMax-M2.5 open source?

MiniMax-M2.5 is released with open weights (Open weights (unrestricted)), so the model can be downloaded and self-hosted.

Benchmark data from Epoch AI (CC BY) and Artificial Analysis; pricing, specs & descriptions from OpenRouter. Data updated Sep 17, 2026.