The past few days have been eventful for the open source AI community, with a new development causing quite a stir. On January 28, a user known as “Miqu Dev” uploaded a set of files on HuggingFace, the leading platform for sharing open-source AI models and code. These files represented what seemed to be a new open-source large language model (LLM) called “miqu-1-70b.”
What caught the attention of many was the similarity between the new LLM’s “Prompt format” and that of Mistral, a well-funded open-source AI company based in Paris. Mistral is known for its Mixtral 8x7b model, which is considered one of the top-performing open-source LLMs available.
Shortly after the files were posted on HuggingFace, an anonymous user on 4chan shared a link to the miqu-1-70b files. Users quickly discovered the model’s exceptional performance in common LLM tasks, rivaling even OpenAI’s GPT-4 on the EQ-Bench test. The news spread like wildfire on social media platforms, with Machine Learning (ML) researchers taking notice on LinkedIn.
