Visit
0 upvotes
MiniMax M2.5 is an open-source frontier model designed for real-world productivity with state-of-the-art performance in coding, search, agentic tool-calling, and office work. It offers efficient execution at $1 per hour with 100 tokens per second, enabling economically viable scaling of long-horizon agents.
MiniMax M2.5 is an open-source frontier model extensively trained with reinforcement learning in hundreds of thousands of complex real-world environments. It achieves state-of-the-art performance across coding, agentic tool use, search, office work, and other economically valuable tasks, with benchmark scores including 80.2% in SWE-Bench Verified, 51.3% in Multi-SWE-Bench, and 76.3% in BrowseComp.
The model demonstrates substantial improvements in programming evaluations, especially in multilingual coding tasks across over 10 languages including Go, C, C++, TypeScript, Rust, Kotlin, Python, Java, JavaScript, PHP, Lua, Dart, and Ruby. It covers the entire development lifecycle from system design and environment setup to feature iteration, code review, and system testing for full-stack projects spanning Web, Android, iOS, and Windows platforms. In search and tool calling, M2.5 achieves industry-leading performance on benchmarks like BrowseComp and Wide Search, with better decision-making and approximately 20% fewer rounds compared to previous versions.
M2.5 was trained to produce truly deliverable outputs in office scenarios through collaboration with senior professionals in finance, law, and social sciences. It shows significant capability improvements in high-value workspace scenarios such as Word, PowerPoint, and Excel financial modeling. The model's efficiency comes from native serving at 100 tokens per second, optimal task decomposition, and improved token efficiency, resulting in 37% faster completion of complex tasks like SWE-Bench Verified.
The model enables innovative agentic applications through cost-effective pricing at $1 per hour for continuous operation at 100 tokens per second, with costs dropping to $0.30 at 50 tokens per second. It powers complex agents without cost concerns, with pricing one-tenth to one-twentieth that of competing models like Opus, Gemini 3 Pro, and GPT-5. M2.5 is deployed in MiniMax Agent, where it handles tasks autonomously using standardized Office Skills and domain-specific Experts for scenarios like office work, finance, and programming.
Target users include developers and organizations needing efficient AI capabilities for coding, search, and office automation. The model integrates through API access and supports various platforms including web applications. MiniMax itself uses M2.5 for 30% of daily operational tasks across R&D, product, sales, HR, and finance functions, with generated code accounting for 80% of newly committed code.
Publisher
A
admin
Launch Date2026-02-18
Platformweb
Pricingpaid