Nvidia unveiled a new partnership with large language model developer Mistral AI, with the companies set to accelerate development and deployment of a new family of open source models.
Under the partnership, the companies said they will utilize Nvidia’s platforms to deploy Mistral’s recently unveiled family, Mistral 3.
This new range is described by Mistral as open source, multilingual and multimodal, with the family optimized across Nvidia’s supercomputing and edge platforms.
Mistral 3 is built using a mixture-of-experts (MoE) architecture, meaning only the relevant part of the model is activated for a task. The model is pitched by the companies as enabling more efficient and accurate deployment.
Nvidia said it’s pairing its GB200 NVL72 systems with Mistral AI’s MoE architecture will let enterprises deploy and scale large models more efficiently, benefitting from “advanced parallelism” and hardware optimization.
“With 41B active parameters, 675B total parameters and a large 256K context window, Mistral Large 3 delivers scalability, efficiency and adaptability for enterprise AI workloads,” Nvidia said in a blog post.
The companies said the new model is available “everywhere from the cloud to the data center to the edge” as of Dec. 2.
The collaboration builds on existing work between the two companies, including development of the Mistral NeMo 12B language model for chatbots and coding tasks.
Alongside the announcement, Mistral also released nine small language models that it said can help developers run AI “anywhere.”
The models are designed to run on Nvidia’s hardware, including Spark, RTX PCs and laptops, as well as Jetson devices. Developers can access the models through AI frameworks Llama.cpp and Ollama.
The Mistral 3 family is available to researchers and developers, a move the company said helps “democratize” access to frontier-class AI.
The news comes in the same week as Nvidia’s announcement that it has invested $2 billion in chipmaker Synopsys, underscoring the company’s push to strengthen its standing in the AI and computing sector.
