.svg){kind=spacer}
Details
- OpenAI announces Deep Research in ChatGPT now powered by GPT-5.2, rolling out starting February 10, 2026, with additional improvements coming soon.
- New features include connecting to apps in ChatGPT for targeted searches on specific websites.
- Users can track real-time progress of research tasks, interrupt with follow-up questions, or add new sources mid-process.
- Fullscreen reports provide detailed, expansive views of research outputs.
- GPT-5.2, OpenAI's flagship model, introduces adaptive reasoning modes: Instant for quick tasks, Thinking for deeper analysis, and Pro for research-grade intelligence; it excels in complex work like coding, science, and data synthesis.
- Previously, Deep Research used o3 and o4-mini models since its 2025 launch; GPT-5.2 enhances agentic capabilities with better long-context understanding and tool-calling.
Impact
OpenAI's upgrade of Deep Research to GPT-5.2 strengthens its position in the AI agent race, where rivals like Anthropic's Claude and Google's Gemini are also advancing multi-step reasoning and web integration, though OpenAI's real-time tracking and app connectivity appear among the more polished implementations to date. This lowers barriers for complex research by enabling targeted site searches and interruptible workflows, accelerating adoption in professional settings like finance and science where synthesis of vast data is routine. By tying into GPT-5.2's state-of-the-art long-context performance—nearing 100% accuracy on 256k-token retrieval tasks—it addresses key agentic bottlenecks, pressuring competitors to match in reliability for end-to-end tasks. Over the next 12-24 months, expect this to steer R&D toward hybrid agent-human interfaces, boosting funding for on-device inference to rival cloud-heavy approaches while navigating scrutiny on hallucination risks despite web safeguards.
