MiniMax M2.5: benchmarks, price and rank
MiniMax M2.5 from MiniMax is 85th of 123 rated AI models on Spezia's leaderboard, strongest on SWE-bench Verified. It costs $0.30 in, $1.20 out per million tokens.
- 85th of 123 rated models
- Rating 940
- Rated on 6 benchmarks
- Open weights
- Released Feb 12, 2026
- 200,000 tokens of context
Where it is strongest
- 6th of 56 on SWE-bench Verified
Where it falls behind
- 100th of 107 on LMArena text
- 59th of 69 on ARC-AGI-2
- 75th of 98 on LMArena WebDev
Price: $0.30 in, $1.20 out per million tokens.
Its headline results
The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.
- BrowseComp76.3%
Questions whose answers are hidden deep on the web and take many searches and pages to track down.
26th of 56. The maker's own number, from MiniMax, via LLM Stats.
- SWE-Bench Pro55.4%
Long changes to real codebases in several languages, often across many files, including code from private and copyleft projects that models are unlikely to have seen.
33rd of 50. The maker's own number, from MiniMax, via LLM Stats.
- LMArena text1,359
People put a question to two anonymous models and vote for the better answer.
100th of 107. Independent result from LMArena.
- LMArena WebDev1,386
People ask two anonymous models to build the same web page or app and vote for the better one.
75th of 98. Independent result from LMArena.
- SWE-bench Verified80.2%
Real bug reports and feature requests from popular open-source Python projects. The model has to change the code so the project's hidden tests pass.
6th of 56. The maker's own number, from MiniMax, via LLM Stats.
- ARC-AGI-24.9%
Visual puzzles of colored grids, each with a new rule to work out from a few examples.
59th of 69. Independent result from Epoch AI.
Its full profile is in Spezia.
Every result with its source, what it costs per task, what jumped with its release, side by side with any other model, and answers to your questions about when to use it.
Models ranked near it
How the rating works, and every benchmark it uses: the model leaderboard.