Researchers Replicate OpenAI’s Hot New AI Tool in 24 Hours

Hugging Face, an AI developer, recently made headlines by creating an open-source AI research agent that can compete with OpenAI’s latest Deep Research feature in just 24 hours. The company, led by Sam Altman, unveiled this new agent over the weekend, claiming that it has the ability to synthesize large amounts of online information and complete complex research tasks.

Deep Research, developed by OpenAI, enhances existing AI models to provide users with new functionalities. Users can request tasks such as generating competitive analyses on streaming platforms or personalized reports on commuter bikes, which can take anywhere from five to 30 minutes to complete. However, Hugging Face researchers quickly developed a comparable alternative to Deep Research.

In a statement released on Tuesday, Hugging Face explained that while powerful Language Learning Models (LLMs) are available in open-source, OpenAI did not disclose much about the framework behind Deep Research. Determined to replicate their results, Hugging Face embarked on a 24-hour mission to create and open-source the necessary framework. The company introduced an “agent” framework that writes actions in code, resulting in a significant performance improvement.

Although Hugging Face’s Open Deep Research achieved a 55.15% accuracy on the General AI Assistants benchmark, while OpenAI’s version scored 67.36%, there is still room for improvement. Despite this, the rapid development of Hugging Face’s agent in just 24 hours showcases the competitiveness of AI tools in the industry. The exchangeability of AI models has become a significant topic, especially with the emergence of DeepSeek, a Chinese AI startup that disrupted the tech sector with its lean and efficient model called R1.

The competition among AI models has led to innovative strategies like distillation, where reasoning capabilities are created by training AI models on the output of others. This approach has raised questions about intellectual property rights, particularly as AI models become more interchangeable. Notably, researchers at Stanford and the University of Washington developed a rival to OpenAI’s reasoning model for less than $50 worth of cloud compute credits, demonstrating the potential for cost-effective AI solutions.

As industry giants like OpenAI and Meta invest billions in expanding AI infrastructure, the rise of cheaper alternatives like DeepSeek has challenged the status quo. The profitability of these tools remains uncertain, especially as smaller players can quickly replicate and offer similar capabilities for free. The future of AI development is evolving rapidly, with competition driving innovation and affordability in the industry.