OpenAI patents AI system that summarizes complex reasoning steps
OpenAI has filed a patent for a system that helps AI models manage complex, multi-step tasks by buffering their internal reasoning and tool results. The system waits for external data to arrive, then combines it with the AI's thought process to create a concise summary of each step. This summary is then sent to an app or displayed to the user.
In the filing’s own words
A system and method are provided for performing artificial intelligence reasoning with buffered summarization and asynchronous tool-call integration. The method includes receiving an input and mapping the input to a set of input tokens, and processing the tokens using an AI model to generate reasoning tokens and one or more tool calls. Tool-call responses are received, queued, and released together with a subset of buffered reasoning tokens to a summarization model. The summarization model generates a step summary of a multi-step reasoning process based on the released reasoning tokens and tool-call responses. The step summary is returned via an application programming interface (API) or displayed in a conversation-thread user interface. By coordinating asynchronous tool-call execution, reasoning token buffering, and incremental summarization, the method improves latency management and enables structured, context-aware summaries for multimodal reasoning outputs.
What exactly they claim (claim 1)
1. A method for artificial intelligence (AI) reasoning, the method comprising: receiving an input and mapping the input to a set of input tokens; processing the set of input tokens using an AI model to generate reasoning tokens and one or more tool calls; performing the one or more tool calls and receiving tool-call responses therefrom; queueing the tool-call responses; buffering the reasoning tokens; determining when to release the reasoning tokens and the tool-call responses to a summarization model to provide released tool-call responses and a subset of released reasoning tokens; generating, by the summarization model and based on the released tool-call responses and the subset of released reasoning tokens, a step summary of a multi-step reasoning process; and returning the step summary to an application programming interface or for display in a conversation-thread portion of a user interface.
Facts
- Company
- OpenAI (OpenAI OpCo, LLC.)
- Official title
- VISUAL REASONING AND MULTIMODAL CHAIN-OF-THOUGHT BACK-END
- Publication number
- US20260330193A1
- Application number
- 19364606
- Filed
- Oct 21, 2025
- Published
- Oct 22, 2026
- Inventors
- 8
- Claims
- 20
Classified under
G06N 3/0475 Physics & computingMore from OpenAI in AI & software
- OpenAI patents voice system that detects when you stop talking Sep 22, 2026
- OpenAI patents system to catch AI models cheating on tasks Sep 10, 2026
- OpenAI patents method to check AI thinking for cheating Sep 10, 2026
- OpenAI patents system that turns photos into new AI-generated images Sep 10, 2026
- OpenAI patents AI that answers questions using only your company's secure files Sep 17, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.