one year on
Mistral releases Mixtral 8x22B under Apache 2.0
Mixtral 8x22B is Mistral's latest open model, a sparse MoE with 39B active parameters out of 141B, released under Apache 2.0.
Originally published Published on this site
Mistral AI announced Mixtral 8x22B on April 17, its latest open model. It is a sparse mixture-of-experts model that uses only 39B active parameters out of 141B, released under Apache 2.0. Mixtral 8x22B supports a 64K token context window, is multilingual in English, French, Italian, German, and Spanish, and natively supports function calling. The model sets a new standard for performance and efficiency, outperforming other open-weight models on benchmarks including MMLU, HellaSwag, and coding tasks.
One year later — open only if you can handle spoilers
Mistral formally announced Mixtral 8x22B on April 17 under Apache 2.0, and the model went on to become a popular base for fine-tuning. The torrent-first release ritual became a defining brand move for Mistral, though later models would be released through more conventional channels.
Filed
Compiled and edited by Adam Allcock.
The Weekly Replay · free by email
This week, one year ago — every Sunday.
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy