Since DeepSeek challenged OpenAI two weeks ago, the open- vs. closed-source AI competition has shown no signs of stopping.

Also: Why Mark Zuckerberg wants to redefine open source so badly

Just two days after OpenAI announced Deep Research, a new AI agent within ChatGPT that can sift through online sources for you, its open-source counterpart has already emerged.

Hugging Face's Deep Research

On Tuesday, Hugging Face released its equivalent to the new feature. Blatantly dubbed open Deep Research, the alternative uses OpenAI's o1 model and an agentic framework to navigate the web. The open alternative achieved 55% accuracy on the General AI Assistants benchmark (GAIA), a top assessment test for agents, compared to Deep Research's 67%, and ranks in first place for open submissions.

However, Hugging Face acknowledged the agent is not yet a full competitor to OpenAI's. "Deep Research is a massive achievement and its open reproduction will take time," the developer platform said in a blog titled "Freeing Our Search Agents." "In particular, full parity will require improved browser use and interaction like OpenAI Operator is providing, i.e. beyond the current text-only web interaction we explore in this first step."

Also: Dumping open source for proprietary rarely pays off: Better to stick a fork in it

OpenAI's Deep Research is underpinned by a version of its latest and most advanced reasoning model, o3, of which there is currently no known open-source equivalent. According to OpenAI's blog, this model version also outperformed top models on Humanity's Last Exam , a new AI benchmark test released just last week, and is much more challenging than other popular tests, with a "new high" of nearly 27% accuracy.

That said, HLE's creators point out a potential "contamination": o3 was evaluated after HLE was released, meaning OpenAI had access to its prompts. Hugging Face did not mention whether it had tested open Deep Research on HLE. To better compete, the platform says it's building "agents that view your screen and can act directly with mouse & keyboard."

It's free to try

Considering its$200-per-month price tag via ChatGPT Pro, Deep Research may be inaccessible to most. If you want to try something similar for free, check out open Deep Research's live demo here, which Hugging Face refers to as a "simplified version" of the full agent.

Also: Are ChatGPT Plus or Pro worth it? Here's how they compare to the free version

The pace at which Hugging Face was able to create something of a competitor -- under 24 hours -- marks the race that makers of proprietary models increasingly find themselves in. Researchers at UC Berkeley made a model comparable to o1-preview in just 19 hours earlier last month. DeepSeek's exact timeline on R1, its o1 rival model, is unknown, but it is understood to be lower-resource in terms of time and spend.

Artificial Intelligence

The best AI for coding in 2025 (and what not to use - including DeepSeek R1)

I tested DeepSeek's R1 and V3 coding skills - and we're not all doomed (yet)

How to remove Copilot from your Microsoft 365 plan

How to install an LLM on MacOS (and why you should)

• The best AI for coding in 2025 (and what not to use - including DeepSeek R1)
• I tested DeepSeek's R1 and V3 coding skills - and we're not all doomed (yet)
• How to remove Copilot from your Microsoft 365 plan
• How to install an LLM on MacOS (and why you should)