A high-performing, industry-standard 7.3B parameter model, with optimizations for speed and context length.
An improved version of Mistral 7B Instruct, with the following changes: