Mistral is launching two new LLMs tailored to run locally on devices. Ministral 3B has three billion parameters, Ministral 8B has eight billion.

Mistral announces in a blog Les Ministraux To. The French AI company’s latest LLM range consists of two models. The name Small language models is perhaps more appropriate here, because Mistral has deliberately kept its models smaller so that they can fit on a laptop or smartphone.

As usual, the name of the models refers to the number of parameters. Ministerl 3B has three billion parameters. If you want something more, Mistral also offers the Ministral 8B model with eight parameters. According to Mistral, the models perform better than Mistral 7B in all benchmarks and can also more than hold their own against Google’s Gemma 2 (2B) and Meta’s Llama 3.1 (8B) and Llama 3.2 (3B).

Tailored to devices

Mistral is clearly following trends in the AI ​​industry. The Ministral models were not designed for a huge data center, but were deliberately made smaller so that they can run locally on a laptop or smartphone without an intermediate step to the cloud. Mistral says there is increasing demand for it from its customers due to privacy arguments. It helps better to sell AI PCs and smartphones.

Both Ministral 3B and Ministral 8B can process up to 128,000 tokens per prompt. The context window is therefore more or less equivalent to OpenAI’s GPT-4 Turbo.

Mistral attracted attention a year ago with the launch of the Mistral 7B, a model that was offered completely free of charge. Since then, the range has expanded rapidly. In September, the company launched its first model with multimodal features: Pixtral 12B. As a European player, Mistral can have a seat at the table with American AI leaders like OpenAI, Google and Anthropic.