Hugging Face, a New York City-based startup known for its developer-focused repository of open source AI code and frameworks, has recently introduced a game-changing feature – third-party customizable Hugging Chat Assistants. This new offering allows users of Hugging Chat to create their own personalized AI chatbots with specific capabilities, similar to OpenAI’s custom GPT Builder. However, unlike OpenAI’s solution which requires a paid subscription, Hugging Chat Assistants are completely free.
According to Phillip Schmid, Hugging Face’s Technical Lead & LLMs Director, building a new personal Hugging Face Chat Assistant can be done “in 2 clicks!” In his announcement on the social network X, Schmid openly compared the new capabilities with OpenAI’s custom GPTs. The major difference, aside from the free access, is that OpenAI’s tools entirely rely on their proprietary large language models (LLMs) GPT-4 and GPT-4 Vision/Turbo. On the other hand, users of Hugging Chat Assistant have the freedom to choose from various open source LLMs to power the intelligence of their AI Assistant, such as Mistral’s Mixtral or Meta’s Llama 2.
The Power of Choice
This flexible approach aligns with Hugging Face’s commitment to providing a diverse range of models and frameworks for users to select from. Just as Hugging Chat offers multiple open source models to power its functionality, Hugging Chat Assistants allow users to tap into different open source LLMs to enhance their chatbot experience.
In a move reminiscent of OpenAI’s GPT Store, Hugging Face has also created a central repository of third-party customized Hugging Chat Assistants. Users can now browse through and choose from a variety of Assistants that best suit their needs. The Assistants page shares a striking resemblance to the GPT Store, with each Assistant presented in rectangular, baseball card-style boxes adorned with circular logos.
Why Hugging Chat Assistants Are Making Waves
Already, members of the open source AI community are praising Hugging Chat Assistants as a superior alternative to GPTs. Mathieu Trachino, founder of enterprise AI software provider GenDojo.ai, expressed his enthusiasm on X, highlighting the merits of the underlying models’ customizability and the fact that the entire solution is free, unlike the subscription-based tiers of OpenAI. Trachino did, however, acknowledge some areas where custom GPTs still outperform Hugging Chat Assistants, such as web search support, retrieval augmented generation (RAG), and the ability to generate logos using OpenAI’s image generation AI model, DALL-E 3.
The introduction of Hugging Chat Assistants demonstrates the rapid progress of the open source community, closing the gap with closed rivals like the ironically named “Open” AI. This milestone comes hot on the heels of a confirmed leak of a new open source model, Mistral’s Miqu, which nearly matches the performance of the closed GPT-4, a benchmark for large language models. The question remains: for how long will closed AI solutions maintain their dominance in the face of relentless open source innovation?
