MiniMax shipped a million-token coding model nobody can download.
The Chinese lab MiniMax released a coding model called M3.1-Flash-Preview on Sunday. It reads a million words at once and offers five levels of reasoning, from low to max. It runs only inside the company's own app, with no model card, no price and nothing to download.
Why it mattersMiniMax has open-sourced earlier models, and this one it kept. Chinese labs are starting to lock their best work inside products, like American ones.
Read in the app — free, in 9 languages
Related stories
Fireworks trained a model to think 40% less for the same answers.2026-09-28Amazon now sells a way to watch what your AI agent actually did.
2026-09-28Google says Gemini 4 has entered its last stage before release.
2026-09-28A startup three months old beat OpenAI on a robot benchmark.
2026-09-27Meituan's 1.6-trillion-parameter model learned to read screenshots.
2026-09-26