Highlights
- GPT-6.2 Prompt Caching: GPT-6.2 models don't cache a system prompt sent in the Responses
instructionsfield, so every step re-billed it as uncached input. OpenCode now sends it as a leading developer message on every Responses route, including OpenAI (API key and ChatGPT sign-in), GitHub Copilot, Zen, Azure and OpenRouter, so it is cached from the second step on. (#64542, #64544)
Core
Bug fixes
- Fixed GPT-6.2 models re-billing the system prompt as uncached input on every step. On GitHub Copilot
gpt-6.2-sol, uncached input per step dropped from about 14,000 tokens to 3. Compaction through/responses/compactnow reads the same cached prefix. (#64542, #64544) webfetchstructured output no longer repeats theurlandformatit was called with. (#64534)