The news, 365 days behind — on purpose Delayed live · replaying 2025

One Year Ago.AI

Remember how fast this is.

10APR2024replayed
one year on
open weightsMistral AI

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.

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.

Replay thisPost on XRedditHNLinkedIn