Better prompt caching for GPT-6
Summary
GPT-6 introduces an improved prompt caching system (a technique that stores repeated instructions and context to avoid reprocessing them) that achieves higher cache hit rates by default and offers developers up to 90% discounts on cached input tokens. The update includes new monitoring tools like the Prompt Caching Dashboard and diagnostics tool to help developers track cache performance and identify why cached content isn't being reused. Developers can also optimize caching by explicitly choosing which prompt sections to cache, adjusting reasoning effort without breaking the cache, and prewarming (loading context ahead of time) the cache to reduce wait times.
Solution / Mitigation
OpenAI provides several explicit mitigation strategies: (1) Use the Prompt Caching Dashboard to monitor cache hit rates and compare cached versus uncached tokens; (2) Use the prompt caching diagnostics tool to compare requests and identify changes to the model, tools, settings, or input that prevented cache reuse; (3) Set explicit cache breakpoints to choose which prompt prefixes to reuse; (4) Adjust reasoning effort using configuration_update instead of removing it entirely to preserve cache; (5) Keep tool definitions, schemas, and ordering stable, using allowed_tools or tool_choice instead of removing definitions; (6) Append new instructions as developer messages rather than modifying earlier ones; (7) Prewarm the cache by preparing known context during application startup before user requests arrive.
Classification
Affected Vendors
Related Issues
Original source: https://openai.com/index/better-prompt-caching-for-gpt-6
First tracked: September 22, 2026 at 08:01 PM
Classified by LLM (prompt v3) · confidence: 95%