MiniMax: MiniMax M3

MiniMax M3 is a commercial model from MiniMax that accepts text, image, and video input and returns text. It offers a context window of 1,048,576 tokens and allows up to 512,000 completion tokens, so it can take in very large documents or long video-derived inputs and produce lengthy outputs in a single call. It supports tool calling and a reasoning mode; structured output support is not listed, so verify that before relying on schema-constrained responses. Pricing is $0.30 per million input tokens and $1.20 per million output tokens, which puts it at the low-cost end for teams processing high volumes of long-context or multimodal material. Its benchmark evidence is thin: a blended score of 45.3 rests on a single independent benchmark (AI Index score of 48.2). Shortlist it if cost, context size, and video input matter most, but test it on your own tasks before committing.