Mistral Previews Open-Weight Sparse Model Family, Sets July Early Access
Mistral said it plans to release a new model this summer as the first in a new sparse model family, with an early-access program starting in July for selected partners in research, government and industry. The company said the new model and subsequent releases will be open-weight.
Mistral said it is pairing the model rollout with portable training and deployment products, Forge and Studio, that can run in a customer’s VPC or datacenter, or on infrastructure it controls that is decoupled from US service providers. The company said Forge can continuously train models using recorded human-AI interaction, which it said should improve efficiency and help companies and governments keep AI systems operating outside external control.
From the sources (4 posts)
@arthurmenschFirst, we have a nice model coming this summer – we hope it will delight and surprise in a few capabilities. This will be the start of a new family of models, fat indeed, but sparse. We're opening up an early access program in July for key
@arthurmenschThis model and upcoming ones will be open-weight. We believe this is critical for our customer confidence and for the research and developer communities. You cannot own, inspect, audit, or improve a system you are only permitted to reach th
@arthurmenschWe've built Studio (for deployment) and Forge (for training) as portable products, and are now hosting them on infrastructure we control. We'll run in your VPC, your datacenter, or on our infrastructure that is decoupled from US service pro
@arthurmenschWe're working with companies and governments around the world to make sure their AI systems are up and running outside of external control, improving with each model release, and with an efficient cost structure. Forge allows to continuousl