Minimax M 2.5

New
#3SWE Bench Verified

by MiniMax

+
+
+
+
About

MiniMax M2.5 is a large language model from MiniMax developed with extensive reinforcement learning training across complex real-world tasks — a training approach designed to improve the model's ability to reason through and complete difficult practical tasks rather than just answering knowledge questions. It reflects MiniMax's investment in RL-driven model development as a path to strong task completion in applied, enterprise-oriented settings.

+
+
+
+
Pricing Range
Input (per 1M)$0.30 -$0.30
Output (per 1M)$1.20 -$1.20
Providers1
+
+
+
+
Timeline
ReleasedFeb 12, 2026
+
+
+
+
License & Family
License
MIT
Performance Overview
Performance metrics and category breakdown

Overall Performance

2 benchmarks
Average Score
59.9%
Best Score
80.2%
High Performers (80%+)
1

Performance Metrics

Max Context Window
204.1K

Top Categories

Coding
59.9%
+
+
+
+
All Benchmark Results for Minimax M 2.5
Complete list of benchmark scores with detailed information
SWE Bench Verified
Coding
80.20
80.2%
Unverified
SWE-rebench
Coding
39.60
39.6%
Unverified