Yan Junjie's company launched M3.1 Flash on October 19th; four days later, it opened the weights and added hosted production inference through Together AI.
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
Primary source: [MiniMax](https://x.com/MiniMax_AI/status/2113452026902589732)
Why it matters #
MiniMax is extending M3.1 Flash beyond its own API: downloadable weights widen developer access, while Together AI's hosted inference offers a production route without customer-managed GPUs.
MiniMax opened the weights for M3.1 Flash on October 23rd, pairing the release with production inference through Together AI so developers can use a hosted service instead of managing GPUs themselves.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
source & further reading
runtimewire.com — original article
Microsoft trains Qwen3.8-27B on 187 tasks to improve agent repeatability
Asuri turns its founders' internal AI agents into a governed workflow product
Rust off Rails calls 60 Markdown files a web framework and lets AI write the code