You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
GPT-6 Astra Ultrafast is listed as an available model in OpenCode 1.18.34 (TUI), but every request immediately fails with:
invalid openai provider options
The debug log shows that OpenCode sets:
"serviceTier": "ultrafast"
but the bundled OpenAI provider rejects it:
Invalid option: expected one of
"auto" | "flex" | "priority" | "default"
path: ["serviceTier"]
openai/gpt-6-astra and openai/gpt-6-astra-fast work with the same OAuth account. Ultrafast also works for this account in the official Codex client.
Expected behavior
openai/gpt-6-astra-ultrafast should work since OpenCode already exposes the model and maps it to serviceTier: "ultrafast".
It looks like the bundled OpenAI provider/AI SDK has an older serviceTier validation schema that does not yet accept "ultrafast".
Activity
github-actions commentedon Oct 6, 2026
This issue might be a duplicate of an existing issue. Please check:
serviceTiersupport was rejected or stripped by the bundled OpenAI SDK; this report describes theultrafasttier being rejected by its validation schema.timwkosatec commentedon Oct 6, 2026
Looks like opencode still uses @ai-sdk/openai 3.0.88 while ultrafast support was added on on 3.0.123
rekram1-node commentedon Oct 8, 2026
merged patch for v1, but recommend u update to v2:
https://opencode.ai/v2/docs