Hot on the heels of the Gemini launch, Google barely had time to revel before the spotlight swung swiftly to MistralAI. The Paris-based startup captured the conversation after announcing its foundational model Mixtral 8x7B.
Mixtral 8x7B, a high-quality sparse mixture of expert models (SMoE), excels in performance and efficiency, outpacing Llama 2 70B with six times faster inference. As the most potent open-weight model under an Apache 2.0 license, it offers an optimal cost-performance balance.
Superior to or on par with GPT-3.5 in standard benchmarks, Mixtral can process up to 32k tokens and supports multiple languages, including English, French, Italian, German, and Spanish. It demonstrates exceptional code generation abilities and can be fine-tuned into an instruction-following model, achieving an impressive 8.3 score on MT-Bench.
Keep reading with a 7-day free trial
Subscribe to Sector 6 | The Newsletter of AIM to keep reading this post and get 7 days of free access to the full post archives.