Hugging Face researchers aim to build an ‘open’ version of OpenAI’s deep research tool

The co-founder and chief scientist of the organization Thomas Olph with a group developer of the AI ​​Dev platform said that they created an “open” version OpeneThe

Deep research to unveil an event on Sunday, crawling the web to compile the research report on any topic. Although fascinating, deep research is currently available only for users to subscribe to subscribed to the Pro-planning of $ 200-one month of Open.

The hug face team project, which they are calling Open Deep Research, consist of an AI model – OpenAI’s O 1 – and an open source “Agent Framework” that helps the model plan to plan its analysis and guides it to use equipment like search engines. O 1 is a owned model (ie gates behind the provided API), but the party says it provides better performance than the “open” model as it is DIPSECR R1The

In less than 24 hours, researchers were able to use a common, text-based browser and a “text inspector” toolkit to read files across the web. Open Deep Research can automatically navigate the web, the team has said that scrolling through pages, manipulating the files and even counting data.

A criteria for General AI Assistants earn a 54%score in Open Deep Research in Guy. This is compared to the Open Deep Research score 67.36%.

I’ve tried Open Deep Research Demo The team has set up – but it couldn’t work. The page was in a heavy understanding when the publication was published; After 10 minutes, it gives an error message spit.

However researchers say they are committed to improve experience, and have created it The source code is available at Githab For inspection and response.

There is that there is that noticeable A number Opina deep research “reproducible” The web, which depends on the open model and tooling. They – and Open Deep Research – Lack – Lack of and 3, the model depends on deep research.

A few, if any, beat the models O3 In criteria related to complex questions and answers to collecting information. Short, deep research options of an open model to compete O3, the real thing cannot measure enough.