Hugging Face Launches HuggingChat v2 ‘Omni’ — A Unified Interface with 115 Open Models and Smart Auto-Routing

Hugging Face has released HuggingChat v2, branded as HuggingChat Omni — a unified chat interface that supports 115 open-source models pulled from 15 different providers, including Groq, Cerebras, Together AI, Novita AI, and more.

Omni aims to streamline user experience by implementing automatic model selection: the system intelligently picks the best model for each prompt behind the scenes, removing the need for users to choose models manually. The routing logic is built on a policy-based approach, with contributions from Katanemo’s Arch-Router-1.5B model for small-scale routing.

HuggingChat Omni is powered by the Hugging Face Inference Providers network — allowing users to tap into hundreds of AI models via top-tier inference backends.

Upcoming enhancements include Model Context Protocol (MCP) support with web search, file upload capabilities, routing improvements, and customizable policy rules for expert users.

"Omni uses a policy-based approach to model selection (after experimenting with different methods). Credits to Katanemo for their small routing model: katanemo/Arch-Router-1.5B," Julien Chaumond, CTO at Hugging Face, said.

Notably, this relaunch comes after Hugging Face previously announced the temporary discontinuation of the original HuggingChat interface in mid-2025. The company cited the decision as a strategic move to build a more integrated and scalable chat experience within its broader ecosystem.

HuggingChat v1, first launched in April 2023, served as an open-source alternative to ChatGPT, allowing community models to be run under one interface. It supported a range of models including LLaMA, OpenAssistant, and others, and experimented with transparency, privacy, and inference optimisations.

Over its lifespan, the platform enabled thousands of users to test and compare models, driving innovation in multi-model chat systems in the open-source AI space.