Reasoning Tied to Conversation History in the Responses API
With GPT-6.2, the API now drops saved reasoning whenever you edit earlier messages, which can change cost and cache hits for agents.
Details
- Each reasoning item is now tied to the conversation history before it, so editing an earlier message makes the API drop the reasoning that follows
- OpenAI says this gives better results with fewer tokens and more cache hits when history stays stable
- Changing reasoning effort or adding tools can cause unintended prefix edits, depending on how the change is applied
- A new input_transformations object in the Responses API helps developers detect those edits
- Applies starting with the newest frontier models, including GPT-6.2
Best for
Long-running agents that build on earlier reasoning across many turns
Use it when
Your app edits earlier messages, changes reasoning effort, or adds tools in the middle of a conversation
Who benefits
Developers running agents on GPT-6.2 in the Responses API
Join the newsletter to get ahead with AI.
I share strategies, new modules, free guides, and tools to get you better at using AI in your work.
Leading a team? See team programs →