Mistral

Mistral AI and NVIDIA launch Mistral NeMo 12B, a 12-billion-parameter language model for enterprise use, excelling in diverse tasks and easy customisation.

Mistral AI releases Mathstral, a STEM-focused language model built on Mistral 7B. It excels in mathematical reasoning, achieving top performance on MATH and MMLU benchmarks for its size. The model is available for use and fine-tuning.

Codestral Mamba offers linear time inference and infinite sequence handling. It excels in code generation and reasoning, matching top transformer models.

On June 5, 2024, the Mistral AI team announced three different entry points for specialising their AI models:

To strike a balance between openness and sustainability, Mistral AI has introduced the Mistral AI Non-Production License (MNPL). This new licence allows developers to use Mistral AI's technology for non-commercial purposes and to support research work.

Codestral is an open-weight generative AI model specifically designed for code generation tasks. It is trained on a diverse dataset of over 80 programming languages.