Tools

Explore

Videos Channels Figures

Atmrix

Tools

Explore

Videos Channels Figures

Atmrix

SC

Sequoia Capital Cast

02/25/25

@ Sequoia Capital

The model that powers Deep Research is a fine-tuned version of the O3 model, specifically trained on challenging browsing and reasoning tasks, enhancing its online search and analysis capabilities.

Video

OpenAI’s Deep Research Team on Why Reinforcement Learning is the Future for AI Agents

@ Sequoia Capital

OpenAI’s Deep Research Team on Why Reinforcement Learning is the Future for AI Agents

02/25/25

Related Takeaways

Isa Fulford

05/08/25

@ Sequoia Capital

The 03 model is effective at searching because it is trained with the same tools and browsing datasets used for deep research.

Isa Fulford

05/08/25

@ Sequoia Capital

We trained a version of GPT-3 specifically for web browsing and data analysis to enhance the capabilities of Deep Research.

Isa Fulford

05/08/25

@ Sequoia Capital

To develop Deep Research, we created reinforcement learning tasks to teach the model browsing and data analysis capabilities, and provided it access to tools for training.

Sequoia Capital Cast

02/25/25

@ Sequoia Capital

Deep Research was developed to tackle tasks requiring extensive online research and reasoning, enabling users to efficiently gather information and make informed decisions across various domains.

Sequoia Capital Cast

02/25/25

@ Sequoia Capital

The quality difference among Deep Research products is evident, stemming from the effort put into constructing the datasets and the optimization capabilities of the O Series models.

Isa Fulford

05/08/25

@ Sequoia Capital

Deep research is designed to be the most comprehensive tool for search queries, while the 03 model is better suited for medium-level searches.

a16z Cast

02/07/25

@ a16z

The reasoning model of DeepSeek allows for deeper answers by thinking through the steps of a problem, which was not released with previous models like OpenAI's.

Y Combinator Cast

02/06/25

@ Y Combinator

DeepSeek's R1 model builds on V3 by applying various algorithmic improvements to enhance its reasoning capabilities, achieving performance similar to OpenAI's models on complex reasoning benchmarks.

Lex Fridman Cast

02/03/25

@ Lex Fridman

The release of the OpenAI o3-mini reasoning model aligns with expectations, but DeepSeek-R1 offers similar performance at a lower cost and reveals its reasoning process, unlike o3-mini.