Home/Models/MiniMax M2.5
minimax logo

MiniMax M2.5

by minimax · Released Feb 2026

Open Source
61.7
avg score
Rank #62
Compare
Better than 73% of all models
Context
197K tokens (~98 books)
Input $/1M
$0.15
Output $/1M
$1.15
Type
text
License
Open Source
Benchmarks
21 tested
Data updated today
About

MiniMax-M2.5 is a SOTA large language model designed for real-world productivity. Trained in a diverse range of complex real-world digital working environments, M2.5 builds upon the coding expertise of M2.1...

Tested on 21 benchmarks with 55.1% average. Top scores: Chatbot Arena Elo — Overall (1404.4%), Chatbot Arena Elo — Coding (1396.3%), OpenCompass — IFEval (91.1%).

Looking for similar performance at lower cost?
gpt-oss-20b (free) scores 61.0 (99% as good) at $0.00/1M input · 100% cheaper
Capabilities
coding
59.5
#41 globally
reasoning
44.4
#57 globally
math
81.8
#17 globally
knowledge
51.6
#92 globally
agentic
6.2
#31 globally
language
67.8
#74 globally
Benchmark Scores
Compare All
Tested on 21 benchmarks · Ranked across 7 categories
Score Distribution (all 233 models)
0255075100
▲ You are here
OpenCompass — LiveCodeBenchV6

OpenCompass Live Code Bench v6. Fresh competitive programming problems to evaluate code generation without memorization.

73.6
LiveBench — Coding

Regularly refreshed coding problems that avoid data contamination. New problems added monthly to prevent memorization.

70.7
LiveBench — Agentic Coding

LiveBench coding tasks that require multi-step reasoning and tool use. Tests planning and execution of complex coding workflows.

51.7
ARC-AGI

Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.

63.7
LiveBench — Reasoning

Regularly refreshed reasoning problems testing logical deduction, spatial reasoning, and analytical thinking.

59.3
LiveBench — Data Analysis

Fresh data analysis tasks testing ability to interpret tables, charts, and statistical data.

49.6
OpenCompass — AIME2025

OpenCompass evaluation on AIME 2025 problems. Tests mathematical reasoning on fresh competition problems.

86.2
LiveBench — Mathematics

Regularly updated math problems that test numerical reasoning, algebra, calculus, and combinatorics.

77.4
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Links
Documentation
Community
BenchGecko API
minimax-m2-5
Specifications
  • Typetext
  • Context197K tokens (~98 books)
  • ReleasedFeb 2026
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.001
Available On
minimax logominimax$0.15
Share & Export
Tweet
MiniMax M2.5 is an open-source text AI model by minimax, released in February 2026. It has an average benchmark score of 61.7. Context window: 197K tokens.