MiniMax M2.5 released: 80.2% in SWE-bench Verified
minimax.io
February 12, 2026
13 min read
Summary
MiniMax M2.5 is a state-of-the-art AI model designed for real-world productivity, achieving scores of 80.2% in SWE-Bench Verified, 51.3% in Multi-SWE-Bench, and 76.3% in BrowseComp. It has been extensively trained using reinforcement learning across hundreds of thousands of complex environments, excelling in coding, agentic tool use, search, and office tasks.
Key Takeaways
- MiniMax M2.5 is trained with reinforcement learning in complex real-world environments and achieves state-of-the-art performance in coding, tool use, and office work tasks.
- The model completes the SWE-Bench Verified evaluation 37% faster than its predecessor, M2.1, and matches the speed of Claude Opus 4.6.
- MiniMax M2.5 costs $1 per hour to run continuously at 100 tokens per second, making it a cost-effective solution for users.
- The model has been trained on over 10 programming languages and can handle the entire development lifecycle, from system design to comprehensive code review.
Community Sentiment
MixedPositives
- MiniMax M2.1 is favored for its speed, cost-effectiveness, and excellent tool calling capabilities, making it a top choice in AI workflows.
- The model is currently free on OpenCode, which enhances accessibility for developers looking to experiment with advanced AI tools.
- The balance of speed, quality, and cost in MiniMax models makes them appealing for various applications, including coding and analysis.
Concerns
- Some users express skepticism about the benchmarks, suggesting they may be too good to be true and questioning the novelty of the training methods.
- Concerns about the limitations of available LLMs in corporate environments highlight the challenges of accessing diverse AI models.
Source
minimax.io
Published
February 12, 2026
Reading Time
13 minutes
Relevance Score
58/100
🔥🔥🔥🔥🔥
Why It Matters
This page is optimized for focused reading: quick context up top, a clean summary block, and a direct path to the original source when you want the full story.