MiniMax M2.7

MiniMax M2.7 is an agent-oriented text model from MiniMax, and it suits teams that want a mid-priced model for tool-heavy coding work. The Hugging Face card lists 229B parameters and shows a licence marked only as other, with terms in a separate file that we have not summarised. It does not state a context length or a release date. OpenRouter lists it from 18 March 2026 with a 196,608-token window, text only, at $0.21 per million input tokens and $0.84 per million output tokens, checked on 10 November 2026. The card recommends a temperature of 1.0 and a top_p of 0.95 for inference. The main catch is the licence: read the file in the repository before you ship anything commercial, because the label alone tells you nothing. Newer MiniMax models also exist. Figures come from the Hugging Face card and the OpenRouter model list. We have not run it ourselves. We would use it as a lower-cost option beside a stronger model, after reading the licence and testing it on our own tasks.