Mistral introduces a new version of its popular 7B Mini language model. Mistral-7B-Instruct-v0.3 is specifically optimized for instruction tasks and offers advanced tokenization and function calling. Mistral provides a Python package called mistral_inference that allows developers to easily interact with the model, including a mistral-chat CLI tool for quick conversations. The model can also be used with the Hugging Face Transformers library. Mistral notes that the model does not currently have any moderation mechanisms. It is meant to demonstrate that the 7B base model can be easily customized.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Sources
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
