Hugging Face has established a two-person team named H4, aimed at the open-source development of AI models similar to ChatGPT. The main research focus of the H4 team is on alignment technology, allowing large language models to learn based on human feedback behavior, relying on dedicated clusters for training. The H4 team has released several open-source large language models, including the conversational Zephyr-7B-α and the improved Falcon-40B. H4 does not directly derive commercial benefits from their tools but accelerates Hugging Face.