French AI startup Mistral unveiled its next generation of Mistral models called Mistral 3.
The Microsoft (MSFT)-backed company said Mistral 3 includes three state-of-the-art small, dense models (14B, 8B, and 3B) and Mistral Large 3 — its most capable model to date — a sparse mixture-of-experts trained with 41B active and 675B total parameters.
“Mistral Large 3 is one of the best permissive open weight models in the world, trained from scratch on 3000 of NVIDIA’s (NVDA) H200 GPUs,” said the company.
Mistral noted that after post-training, the model achieves parity with the best instruction-tuned open-weight models on the market on general prompts, while also showing image understanding and best-in-class performance on multilingual conversations (i.e., non-English/Chinese).
Mistral Large 3 debuted at number two in the OSS non-reasoning models category (number six overall among OSS models overall) on the LMArena leaderboard.
The company added that all its new Mistral 3 models, from Large 3 to Ministral 3, were trained on Nvidia Hopper GPUs to tap high-bandwidth HBM3e memory for frontier-scale workloads.
Mistral 3 is available on Mistral AI Studio, Amazon Bedrock, Azure Foundry, Hugging Face (Large 3 & Ministral), Modal, IBM WatsonX, OpenRouter, Fireworks, Unsloth AI, and Together AI. In addition, it will soon come on Nvidia NIM and AWS SageMaker, the company noted.
Earlier on Tuesday it was reported that Microsoft (MSFT)-backed OpenAI (OPENAI) is developing a large language model dubbed Garlic to counter Google’s (GOOG) (GOOGL) recent gains in AI development.
In September, it was reported that Mistral was in final talks to raise €2B that would value the startup at €12B.