unknown

MiniMax M2.5 is a new 230B parameter mixture-of-experts coding model that rivals Claude Opus 4.6 performance at one-tenth the cost. With open weights, it excels at agent workflows, multi-file refactoring, and tool calling while using 20% fewer tool calls and 5% less token waste. Benchmarks show it scoring over 80% on SWE-bench verified (comparable to Opus) and beating Opus slightly on Droid. Pricing starts at $0.15 per million input tokens versus Opus's $5, making it viable for running AI agents at scale. The model supports local deployment, fine-tuning, and includes a free API tier.

How Is This Almost as Good as Opus?

Development tricks and tips. Radically better observability stack. Ship higher-quality software faster. Be the hero of your engineering teams.