Bringing open AI models to the frontier

Mistral AI announced an open-model-first approach to generative AI company mission and actions.

Mistral 7B released a 7-billion-parameter model made available under Apache 2.0 released under Apache 2.0 license today.

Outperforms all currently available open models up to 13B parameters on standard English and code benchmarks compared to open models up to 13B on standard English and code benchmarks.

Mistral AI team assembled the team, rebuilt an MLops stack, and designed a data processing pipeline in three months work done over three months.

Mistral 7B achieved performance placing it among smallest models exceeding 60% on MMLU benchmarks benchmarks referenced include MMLU performance.