Omnifact now supports the Mistral AI API platform and their family of models.

Omnifact now supports the Mistral AI API platform and their family of models.

Feature Drop: New Models and Mistral AI API Support

Published on December 18th, 2023

Following the release of the Mistral AI API platform, we’re happy to announce that the family of Mistral models is now available to be used within Omnifact.

Who is Mistral AI?

Paris-based AI startup Mistral AI has recently emerged as a key player in the generative AI landscape with their powerful open-source and open-weight models, Mistral 7B and Mixtral 8x7B. Accelerated by their recent $415 million Series A funding round at a soaring $2 billion valuation, Mistral has just entered the LLM provider market with their hosted API platform aptly titled “La plateforme”. Being a European-based provider, Mistral AI has the possibility to serve as a GDPR-compliant alternative to well-known US-based providers, such as OpenAI, Anthropic, and Google.

New Models for Omnifact

With the integration of the Mistral AI API platform, we now have three new additions to the Omnifact model family.

Mistral Tiny

Mistral Tiny is a hosted version of Mistral 7B Instruct v0.2, which serves as a cost-effective English-only model. It is Mistral AI’s smallest and cheapest model, scoring a 7.6 on MT-Bench. Our early experiments have shown that although very capable for its cost and size, Mistral Tiny does have some problems following instructions and gets confused sometimes, so please be cautious when enabling Mistral Tiny for your team.

Mistral Small

The next model is Mistral Small, which is a hosted version of Mistral’s newest model, Mixtral 8x7B. Mistral Small serves as an all-purpose multilingual model with support for English, French, Italian, German, and Spanish. It is capable of most creative and logical tasks, including writing code, scoring 8.3 on MT-Bench.

Mistral Medium

Last but not least, Mistral Medium is Mistral AI’s prototype model currently only available through the Mistral AI API. Being Mistral’s highest-quality offering, this model excels at multilingual and coding tasks, achieving a top score of 8.6 on MT-Bench.

Give it a Try

Starting today, all three of Mistral AI’s models are available to be used within Omnifact, all you need to do is enable the Mistral AI provider in your team settings and provide your API key for the Mistral AI platform. Access to the Mistral AI API is currently in early-access and requires registering for their waitlist, which is available here.

Don’t have access to Omnifact yet? Request a live demo and get started today.

© 2024 Omnifact GmbH. All rights reserved.