MiniMax M2.5
MiniMax's open-weight model extending coding skill into general office work
MiniMax M2.5, released February 12, 2026, extends the coding-focused M2.1 model into general office work: generating and editing Word, Excel, and PowerPoint files and switching between different software environments within an agentic workflow. It’s a mixture-of-experts model with 230 billion total parameters and 10 billion active per token, ships with a 204,800 token context window and up to 128,000 tokens of output, and comes in two variants, a standard mode running at roughly 50 tokens per second and a Lightning mode at double that speed. MiniMax prices API access at $0.15 per million input tokens and $0.90 per million output tokens, among the cheapest rates of any frontier-class model at release, and it ships with open weights under a modified MIT license for self-hosting.