Hugging Face Introduces Open Source AI Assistant Creator to Compete with OpenAI's Custom GPT Models

Hugging Face, the New York City-based startup known for its open source AI code repository (which hosted last year’s “Woodstock of AI”), has launched customizable, third-party Hugging Chat Assistants.

This new, free offering enables Hugging Chat users to easily create personalized AI chatbots with specific features. This functionality parallels OpenAI’s custom GPT Builder, which requires a paid subscription ranging from $20 for ChatGPT Plus to variable rates for the Team and Enterprise plans.

Phillip Schmid, Hugging Face’s Technical Lead and LLMs Director, announced the news on X (formerly Twitter), highlighting that users can set up a Hugging Face Chat Assistant "in 2 clicks!" He also noted the resemblance to OpenAI’s custom GPTs.

A major distinction is that Hugging Chat Assistants are entirely free, while OpenAI’s tools rely on proprietary large language models (LLMs) such as GPT-4. In contrast, Hugging Chat Assistant users can select from various open source LLMs, like Mistral’s Mixtral and Meta’s Llama 2, empowering them to customize their AI’s intelligence.

Hugging Face maintains its commitment to providing a diverse range of models and frameworks, allowing users to choose from multiple open source options within both Hugging Chat and Hugging Chat Assistants.

The company has also established a central repository of these third-party customized Hugging Chat Assistants, resembling OpenAI’s recently launched GPT Store. This aggregator page showcases custom Assistants within rectangular, baseball card-style boxes adorned with circular logos.

Some users within the open source AI community have praised Hugging Chat Assistants as “better than GPTs.” For instance, Mathieu Trachino, founder of GenDojo.ai, highlighted their customizability and the fact that they are free, contrasting with OpenAI’s paid tiers. However, he acknowledged that custom GPTs still have advantages, such as web search support, retrieval augmented generation (RAG), and the ability to create logos using OpenAI’s DALL-E 3.

The introduction of Hugging Chat Assistants underscores the rapid advancements within the open source community, especially following the recent leak of Mistral's new model, Miqu, which approaches the performance of GPT-4. The landscape continues to evolve, raising questions about how long the proprietary models will maintain their lead.