French AI company Mistral released Mistral Large yesterday (26 February), its new language model designed for advanced text generation, available through la Plateforme and Microsoft Azure.
Mistral Large is a text generation model with improved reasoning capabilities, making it suitable for complex multilingual tasks such as text understanding, transformation, and code generation.
Mistral Large is currently the second-ranked model available through an API globally, trailing only GPT-4.
Mistral AI was the centre of generative AI chats at the World Economic Forum in Davos in January 2024.
The Paris-based company was created a mere nine months ago, and it reached a €2bn ($2.2bn) valuation last month after a founding round of €400m.
Unlike its rivals such as OpenAI and Google, Mistral AI believes that its technology should be available as open-source software.
The model introduces several new features and strengths, including native fluency in English, French, Spanish, German, and Italian.
Its 32,000-token context window allows precise information recall from extensive documents. Mistral Large offers precise instruction-following, allowing developers to design moderation policies.
Mistral also announced a partnership with Microsoft to bring its models to Azure. Mistral Large is now available through Azure AI Studio and Azure Machine Learning.
The model is also proficient in coding and math tasks.
Mistral has also launched Mistral Small, a model tailored for low-latency workloads. Mistral Small acts as a midpoint between Mistral’s open-source model and Mistral Large.
