🤖AI Agents Weekly: Deep Research, Gemini 2.0 Flash, Rethinking Mixture-of-Agents, Understanding Reasoning LLMs

Deep Research, Gemini 2.0 Flash, Rethinking Mixture-of-Agents, Understanding Reasoning LLMs

In today’s issue:

• OpenAI releases Deep Research, an agent that performs extensive research

• Google announces Gemini 2.0 Flash GA

• Multi-agent Architecture Search

• Rethinking Mixture-of-Agents

• Agentic Object Detection

• Replit Agent on Mobile

• Open-source AI agent to automate tasks and much more.

---

---

Top Stories

OpenAI’s Deep Research

OpenAI has launched deep research, a new capability in ChatGPT that conducts comprehensive multi-step research tasks using the internet and user-uploaded files. The system is powered by a version of OpenAI's upcoming o3 model optimized for web browsing and data analysis.

Deep research achieves new state-of-the-art performance on several benchmarks, including 26.6% accuracy on Humanity's Last Exam (compared to 9.1% for OpenAI o1) and leading scores on the GAIA evaluation. The system excels at complex tasks like analyzing scientific literature, investigating historical events, and synthesizing information from multiple sources.

The technology works by breaking down research queries into steps, searching and analyzing online sources, and compiling comprehensive reports with citations. It can handle tasks that would take humans several hours, from academic research to market analysis. Key features include the ability to read user-uploaded files, generate data visualizations, and provide detailed citations for its findings.

Access is initially limited to Pro users (100 queries/month), with Plus and Team users gaining access later. OpenAI plans to release a faster, more efficient version and expand access to specialized data sources in the future. The company conducted extensive safety testing and classified deep research as Medium risk, with ongoing monitoring and additional safeguards planned.

Blog | Deep Search Guide