IBM adds Mistral Large language model to watsonx.ai

IBM has made the Mistral Large advanced large language model (LLM) available on IBM’s watsonx.ai enterprise tools platform for AI developers, expanding the collection of models accessible on watsonx.ai.

With Mistral AI, watsonx.ai users gain a generative AI model optimized to handle complex enterprise tasks that require advanced reasoning and multilingual capabilities, IBM said. These include retrieval-augmented generation (RAG) specialization, to support longer chat interactions and large document processing; function calling, for connections to external tools such as user-defined functions; and coding proficiency, for generating and commenting on code, with capability to output results in the JSON format. Also offered are “responsible AI,” with built-in guardrail features, and support for multilingual capabilities, with fluency in English, French, German, Spanish, and Italian.

Mistral Large initially will be available as SaaS, IBM said on July 18. Mistral Large joins a collection of models accessible on watson.ai, including the IBM Granite family of foundation models. Mistral AI and IBM continue to foster innovation and collaboration by providing enterprises with access to a mix of open source and commercial models as well as smaller models on watsonx.ai that prioritize data control, IBM said. IBM will provide customers with protection for Mistral Large via a capped intellectual property indemnity, thus making it the first third-party foundation model that gets IBM indemnification. With Mistral Large now on the watsonx platform, enterprises benefit from additional enterprise-ready products, including data store, prompt lab, model tuning, production monitoring, and governance capabilities, IBM said.