AI BriefingCheaper, Better, Faster, Stronger
Key point
Mistral AI has released Mixtral 8x22B, a new open model that maximizes efficiency and performance, under the Apache 2.0 license.
Details
Mistral AI's latest open model, Mixtral 8x22B, adopts a Sparse Mixture-of-Experts (SMoE) architecture. Out of a total of 141B parameters, it uses only 39B active parameters, delivering overwhelming cost efficiency.
Key features include:
- Support for English, French, Italian, German, and Spanish
- Strong math and coding capabilities
- Native function calling support
- A 64K token context window
The model is distributed under the Apache 2.0 license, allowing anyone to use it without restriction. It is faster than existing dense 70B models while showing better performance than other open-weight models.
In benchmark results, it achieved excellent scores across various metrics including MMLU, HellaSwag, and Arc Challenge, and in particular demonstrated multilingual capability surpassing LLaMA 2 70B.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Qwen3.5-based Small Decision Model 'Kev' Released with TypeSafe API Compatibility and Performance Metrics