OpenAI o3-deep-research
OpenAI · United States · 2025
The first model sold as a researcher rather than a chatbot — and the first to be retired on three months notice instead of six.
o3-deep-research reached the API on 26 June 2025 as OpenAI's most powerful model for multi-step research. It was not sold as a conversational model at all. Given a question, it would search the web, run code, pull material from private sources through MCP connectors and return a synthesised report with citations — a single API call that could run for many minutes and issue dozens of tool calls of its own accord. The specification followed the o-series: a 200,000-token context window, up to 100,000 output tokens, text and image input, text output, and a knowledge cutoff of June 2024. Pricing was 10 dollars per million input tokens, 2.50 for cached input and 40 for output. Its supported tool list was deliberately narrow — web search, code interpreter and MCP — because the point of the model was the research loop, not general function calling. Its retirement is the more instructive part of the story. OpenAI classifies deep research models as specialised variants of generally available models, a category that carries a minimum notice period of three months rather than the six months granted to general models. The deprecation was announced on 22 April 2026 and the model shut down on 23 July 2026, with gpt-5.6-sol as the replacement. Deep research is now a tool configuration on a general model rather than a product of its own.
▸News
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!