Julien Chaumond, CTO of Hugging Face, announced today the closure of HuggingChat, the company’s experimental open source chatbot interface, marking the end of a notable chapter in the evolution of community-driven AI applications.
Launched in April 2023, HuggingChat emerged just five months after the release of ChatGPT. At the time, deploying a comparable conversational AI service was still technically challenging. HuggingChat set out to prove it was possible to build an open alternative leveraging the fast-growing ecosystem of community-trained models and modular tools.
Beyond offering a free chatbot, the project played a pivotal role in advancing inference optimization. Work on HuggingChat led to the development of text-generation-inference, which laid the groundwork for contributions to key LLM runtime engines like SGLang, vLLM, and llama.cpp.
Over its 27-month lifespan, HuggingChat introduced several innovative features:
- Assistants, allowing users to bundle a model, prompt, and tools into a shareable link
- HuggingChat Tools, which integrated Gradio Spaces directly into the chat interface
- A long list of supported model launches, including OpenAssistant, all the Llama family models, Phi-4, Qwen, DeepSeek, and Gemma
Chaumond described the decision to sunset the project as “bittersweet,” noting that while HuggingChat has always been free and experimental, the time has come to make space for “something new and more integrated with the HF ecosystem.”
Users can export their past conversations as a zip file to load them into other interfaces.
For those looking to continue experimenting with open source chat UIs, Chaumond recommended a few strong alternatives:
- LibreChat by Danny Avila, a simple but powerful interface
- Open WebUI by Tim Jaeryang Baek, now widely adopted across the open model community
- Scira MCP, an emerging interface by Zaid Mukaddam focused on multi-component workflows
