A high-performing, industry-standard 7.3B parameter model, with optimizations for speed and context length. An improved version of [Mistral 7B Instruct v0.2](/models/mistralai/mistral-7b-instruct-v0.2), with the following changes: - Extended vocabulary to 32768 - Supports v3 Tokenizer - Supports function calling NOTE: Support for function calling depends on the provider.