OpenAI updates GPT-6 prompt caching with new controls
By Automate Basics, written with AI from OpenAI's original
OpenAI says GPT-6 is more likely to reuse cached prompt material and adds diagnostics, explicit breakpoints and controls aimed at reducing delays and costs. The announcement is relevant to teams that send similar prompts repeatedly, but gives no plan or pricing details.
What changed in GPT-6 prompt caching?
OpenAI says GPT-6 is more likely to find a usable cache match when processing prompts. It also describes new diagnostics, explicit breakpoints and controls intended to reduce response time and cost. Prompt caching matters when work requests contain material that stays the same, such as recurring instructions alongside a changing question.
The available announcement summary does not explain how the diagnostics work or what a breakpoint changes. It also does not quantify the claimed improvement. For now, the practical news is that OpenAI has announced more ways to inspect and manage caching, not that every repeated prompt will produce a measurable saving.
Who can use the new caching controls?
OpenAI does not say which plans include the GPT-6 caching changes, when users can access them or what they cost. The announcement summary also does not say whether the controls appear in a workplace app, a developer tool or both. Teams should not assume that a ChatGPT subscription includes them.
The changes are most relevant to people whose work sends similar instructions to GPT-6 many times, especially if delays or usage costs matter. Someone who asks occasional, unrelated questions may have less reason to investigate caching. Before planning around the update, a team would need to confirm access and pricing through information OpenAI has not provided in the available summary.
How do I try GPT-6 prompt caching at work?
A sensible first task is a recurring request with shared instructions and changing input, such as sorting similar incoming messages into categories. If your team can access the new GPT-6 controls, use a task you already understand rather than redesigning a workflow around an unverified saving.
Check whether the answers still follow the shared instructions and handle each new message correctly. If your existing tools show response times or costs, compare those records before and after any change. OpenAI says the update aims to reduce both, but the available summary gives no procedure, expected saving or explanation of what the new diagnostics display.
Frequently asked questions
What did OpenAI announce for GPT-6 prompt caching?
OpenAI says GPT-6 should find usable cached prompt material more often. It also announced diagnostics, explicit breakpoints and controls aimed at reducing response time and costs. The available summary does not explain how those features work or provide a measured improvement.
Is GPT-6 prompt caching included in my plan?
The available OpenAI summary does not identify eligible plans, give a price or state when the changes become available. It also does not establish whether they are included in ChatGPT. Check those details before counting on the controls for a work process.
Related reading
Google tests Flipkart checkout inside Gemini and AI Mode in IndiaGoogle is testing purchases from select Flipkart listings inside Gemini and AI Mode in India, TechCrunch reports. The limited test may interest teams that use AI to research work purchases, but most users cannot try it yet.
Meta outlines Mac computer use and connectors for Muse AI appMeta announced plans to add Mac computer use, connectors and a dedicated email address to its Muse AI app, according to TechCrunch. The features could make the personal agent more useful for work, but Meta has not said when they will arrive.
Meta's Muse now offers a browser and download for its cloud filesMeta's Muse now offers users a file browser and a downloadable archive of the files in its cloud computer, according to The Verge. The change gives people using Muse for work a more direct way to inspect that environment.
Written with AI from OpenAI's original and published after automatic checks: every figure here appears in the original, and no sentence is copied from it. The picture is AI-generated. The original is the authority.
Source: Better prompt caching for GPT-6, OpenAI, 22 Sept 2026.