IBM: Mistral AI's LLM model integrated into watsonx

(CercleFinance.com) - IBM announced on Thursday that watsonx, its AI and data platform, would henceforth integrate Mixtral 8x7B, the open-mode large language model (LLM) developed by French start-up Mistral AI.


In a press release, the American technology group explains that Mixtral-8x7B was able to process almost 50% more data than other existing models in tests carried out in-house.

IBM explains that the integration of Mixtral-8x7B into its range of tools is part of its strategy to offer its customers several options to help them deploy their AI applications.

The company plans to integrate Elyza-japanese-Llama-2-7b, another open source LLM model developed by Elyza of Japan, into watsonx.

Meta's Llama-2-13B-chat and Llama-2-70B-chat models, also available in open source, are also expected to be added to the watsonx platform in the coming months.

By making the parameters of its models public in open mode, Mistral AI aims to enable institutions, researchers and companies to test and improve them, but above all to readapt them to their particular use cases and contexts.

Copyright (c) 2024 CercleFinance.com. All rights reserved.