Mistral introduced a new open-source AI language model

Paris-based startup Mistral AI has developed the open-source Large Language Model (LMM) Mistral 7B. It is positioned as a more powerful competitor to big players like OpenAI’s ChatGPT.

According to the developers, the new technology is ahead of its rivals in almost all respects.

Mistral’s main emphasis is on the openness and adaptability of the tool.

“We release Mistral 7B under the Apache 2.0 license, it can be used without restrictions. Working with open models is the best way for suppliers and users to build sustainable businesses based on artificial intelligence solutions,” the company emphasized.

This approach helps democratize AI and ensure transparency, while also allowing other teams to test, adapt and improve the existing code base, the developers say.

Mistral provides additional privacy for users because the neural network can be run on local computers and does not require access to the Internet or certain cloud services.

The LMM code is available in the HugginFace repository.

“Mistral 7B demonstrates that the language model can be easily tuned to achieve high performance. But it does not have a moderation mechanism. We look forward to working with the community to find ways to force the model to strictly adhere to constraints to deploy it in ecosystems that require moderated output,” the company noted.

Meta previously presented a set of AI tools – the company talked about a new voice assistant, a neural network with different personalities, smart glasses and stickers.

Let us recall that in September Microsoft announced the Copilot system, which combines Windows OS interfaces with artificial intelligence language models.

Found an error in the text? Select it and press CTRL+ENTER