MiniMax releases MiniMax M2.5, built for real-world productivity
Original title:MiniMax M2.5: Built for Real-World Productivity.
AI summary
MiniMax's newly released large language model M2.5 achieves SOTA on tasks including coding, agent tool calling, search and office work, trained via reinforcement learning in hundreds of thousands of complex real-world environments. The model is highly efficient, completing the SWE-Bench Verified evaluation 37% faster than its predecessor M2.1, on par with Claude Opus 4.6. On pricing, running at 100 tokens/s costs just $1 per hour. M2.5 was trained in more than 10 programming languages and over 200,000 real environments, with full-process capability from system design to testing.
Key Takeaways
- MiniMax's newly released large language model M2.
- 5 achieves SOTA on tasks including coding,agent tool calling,search and office work,trained via…
- The model is highly efficient,completing the SWE-Bench Verified evaluation 37% faster than its…
- 1,on par with Claude Opus 4.
- On pricing,running at 100 tokens/s costs just $1 per hour.
Quick Facts
- Published
- Updated
- Source
- MiniMax:Blog(网页)
- Category
- Model Releases
- AI score
- 67
MiniMax M2.5 把 SWE-bench 拉到 80.2%,成本只有 Claude Opus 4.6 的十分之一,速度还翻倍,对做 agent 的团队来说是个高性价比选择。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: MiniMax releases MiniMax M2.5, built for real-world productivity. February 12, 2026. https://new.nat6.net/en/news/cmpunu0g60007slbrqtizmyjw/