Mistral AI, a startup based in Paris, has revealed its alternative to OpenAI and Anthropic by launching Mistral Large, a large language model. The company described Mistral Large as an advanced text generation model with exceptional reasoning abilities in a blog post.
Mistral AI stated that Mistral Large is capable of handling complex multilingual reasoning tasks such as code generation, transformation, and reading comprehension. The company has also introduced its own version of Chat GPT called Le Chat, currently in beta testing phase. Initially, Mistral AI focused on being open-source as a key feature. The first model was released under an open-source license, however, subsequent larger models have not followed suit.
Introducing Mistral Large
Similar to OpenAI, Mistral AI offers Mistral Large through a paid API with pricing based on usage. According to Tech Crunch, Mistral Large is priced at $24 per million output tokens and $8 per million input tokens to utilize the model. Tokens are used to represent small segments of words, typically divided into syllables. For example, the word “ReadWrite” would be segmented into “read” and “write” for individual processing by the AI language model.
Furthermore, Mistral AI inherently supports context windows of up to 32,000 to accommodate over 20,000 English words and various other European languages like Italian, French, German, and Spanish, as per the report.
In addition, Mistral AI has launched Le Chat, its own rendition of Chat-GPT, which can be accessed at chat.mistral.ai and is currently in beta stage.
Users can select from three models: Mistral Large, Mistral Small, and Mistral Next — a prototype designed to be concise and to the point. As of now, Le Chat is available for free usage, but this may change in the future.
Featured Image: Canva
