opencode 1.18.19: Cloudflare AI Gateway passthroughs, Codex rate-limit matching
opencode 1.18.19 adds native OpenAI and Anthropic passthroughs for Cloudflare AI Gateway models and aligns Codex rate limits with ChatGPT subscription limits. The rest is bugfixes across provider auth display, pricing parsing, and websocket handling.
PUBLISHED2026-08-20
OBSERVED2026-08-21
AGE6d
SOURCES1
- Native OpenAI and Anthropic passthroughs for Cloudflare AI Gateway models.
- Codex rate limits matched more closely to ChatGPT subscription limits.
- Removed built-in Qwen sampling defaults that could send unsupported settings.
/connectnow correctly shows authenticated providers.- Malformed model pricing is ignored instead of breaking usage cost calculation.
- Fallback when OpenAI websocket requests exceed message size limits.
- ChatGPT workspace compute residency forwarded to Codex requests.
- Web search enabled when using the OpenCode Go provider.
- Updated default Console URL for account connection flows; compatibility preserved with existing v1 databases.
- Desktop: saved server details stay editable in the server dialog.
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.