Mistral AI

35 summarised stories about Mistral AI, each linking back to the original source. Browse all topics →

Monday, 11 December 2023

La Plateforme

Mistral AI 2 years ago

Mistral AI launched a beta platform offering three text-generation endpoints and one embedding endpoint for developers to access and customize its open models. The endpoints range from Mistral-tiny (7B parameters, scoring 7.6 on MT-Bench) to Mistral-medium (a prototype scoring 8.6 on MT-Bench), with an embedding model achieving 55.26 on MTEB retrieval tasks. Developers can now register for API access with capacity ramping up progressively, allowing integration of Mistral's models into production applications with different performance-cost tradeoffs.

API & Developer Tools Code Generation Embeddings Mistral AI

Mixtral of experts

Mistral AI 2 years ago

Mistral AI released Mixtral 8x7B, a sparse mixture-of-experts language model with open weights under Apache 2.0 license. The model has 46.7 billion total parameters but uses only 12.9 billion per token, matching or outperforming GPT-3.5 and Llama 2 70B on most benchmarks while running 6 times faster. Developers can now deploy the model using open-source tools like vLLM and access it through Mistral's API platform.

Language Models Mistral AI Mixture-of-Experts Open Source