Repository navigation
[FEATURE]:Support GPT-5.6 prompt caching defaults #36318
Description
Feature hasn't been suggested before.
- I have verified this feature I'm about to request hasn't been suggested before.
Describe the enhancement you want to request
OpenAI introduced new prompt caching behavior for GPT-5.6 and later models.
We should make opencode automatically use the right GPT-5.6 cache defaults while keeping GPT-5.5 and older models on the existing behavior.
Why
GPT-5.6 uses prompt_cache_key for more reliable cache matching and supports request-wide cache options through prompt_cache_options.
The GPT-5.6 family includes multiple model ids, such as:
gpt-5.6gpt-5.6-solgpt-5.6-terragpt-5.6-luna
There may also be provider or suffix variants like gpt-5.6-fast, so detection should match the whole gpt-5.6* family rather than only the exact gpt-5.6 id.
Desired behavior
For GPT-5.6 family OpenAI models, opencode should send:
promptCacheOptions: {
mode: "implicit",
ttl: "30m"
}Reacted by wbdb, Maria Olsen, Tomasz Boguszewski, Romnet, Nalin Mehra, Hegyi Áron Ferenc and Bren
Metadata
Metadata
Assignees
Labels
No labels
Activity
github-actions commentedon Jul 10, 2026
This issue might be a duplicate of existing issues. Please check:
prompt_cache_key) — related prompt_cache_key support across providers.rekram1-node commentedon Jul 16, 2026
yeah we will support this in v2
github-actions commentedon Sep 15, 2026
To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.