Repository navigation
[Bug] Desktop 1.18.29 can offer OAuth models missing from active provider state #47490
Description
Correction
The original provider-prefix diagnosis was incorrect. OpenCode intentionally formats ProviderModelNotFoundError as <providerID>/<modelID>, so openai/gpt-6-astra does not prove that the provider prefix was passed inside modelID.
The verified failure is that OpenCode Desktop 1.18.29 offered a current OpenAI OAuth model in the picker, but that model was missing from the running provider state. The unfiltered catalog still contained it, which is why the error suggested the same model ID.
Description
With the built-in OpenAI provider connected through ChatGPT/Codex OAuth, Desktop accepted a prompt but showed no assistant response. The backend failed before streaming:
ProviderModelNotFoundError: Model not found: openai/gpt-5.6-terra.
Did you mean: gpt-5.6-terra, ...?
The same occurred with gpt-6-astra. Both are documented OpenAI models:
The persisted Desktop selection was correct:
{
"providerID": "openai",
"modelID": "gpt-6-astra"
}No OPENAI_* environment overrides or external plugins were configured.
Local recovery
Refreshing the OpenCode model catalog and restarting Desktop repaired the local provider state:
opencode models openai --refreshAfter the refresh, a clean CLI request using the same OpenAI OAuth login and model succeeded:
opencode run --pure --model openai/gpt-5.6-terra \
--format json 'Reply with exactly: OPENCODE_GPT_OK'It returned OPENCODE_GPT_OK. This confirms that authentication, account access, model availability, and the OpenAI transport are healthy.
Expected behavior
Desktop should refresh/rebuild its active provider model state when the catalog changes, or at least surface an actionable error with a refresh/restart recovery path.
Actual behavior
A model present in the picker/catalog can be absent from the running OAuth provider state. Prompt processing fails before streaming, while the Desktop UI appears to do nothing.
Version and platform
- OpenCode Desktop 1.18.29
- Ubuntu 26.04.1 LTS, Linux x86_64
- Electron 42.3.3
Related issues
- [bug] opencode run: unknown model returns a generic 500 UnknownError instead of a 404 with the "did you mean" message #47162 covers model-not-found errors becoming opaque.
- bug: UI stuck on 'thinking' indefinitely after stream error, no error displayed or state recovery #32366 covers Desktop not surfacing/recovering from backend errors.
Activity
github-actions commentedon Sep 5, 2026
This issue does not fully meet our contributing guidelines.
What needs to be fixed:
gpt-5.5-pro,gpt-5.6-terra,gpt-6-astra,gpt-6-astra-pro,gpt-6-astra-fast,gpt-5.6-terra-pro,gpt-5.6-terra-fast) do not appear to be real OpenAI models. Please reproduce with actual, existing model IDs and update the report accordingly.Please edit this issue to address the above within 2 hours, or it will be automatically closed.
This issue might be a duplicate of existing issues. Please check:
ProviderModelNotFoundError/ "did you mean the same ID?" structural patternparseModelbreaks URL-based provider IDs used byopencode auth login#39926:parseModelbreaks on URL-based provider IDs — sameparseModel/getModelresolution failure classIf you believe this was flagged incorrectly, please let a maintainer know.
[-][Bug] Desktop 1.18.29 prefixes OpenAI modelID twice, so all GPT prompts fail silently[/-][+][Bug] Desktop 1.18.29 passes provider-prefixed modelID to OpenAI provider[/+]oussama1598 commentedon Sep 5, 2026
Updated the report to address the compliance warning:
gpt-5.5-proreproduction.gpt-5.6-terraandgpt-6-astra, both listed in the official OpenAI documentation:-pro/-fastnames were emitted by OpenCode's ownDid you meansuggestions and are not model claims in this report.OPENAI_BASE_URLor otherOPENAI_*environment override.The underlying reproduction remains: Desktop persists
providerID: openaiand an unprefixedmodelID, whileSessionPrompt.getModelreceives/reportsopenai/<model>and fails before streaming begins.github-actions commentedon Sep 5, 2026
Thanks for updating the issue! The models you referenced (
gpt-5.6-terraandgpt-6-astra) are confirmed real OpenAI models, and the report now includes clear reproduction steps, log excerpts, and system information. Everything looks good — the compliance flag has been removed.[-][Bug] Desktop 1.18.29 passes provider-prefixed modelID to OpenAI provider[/-][+][Bug] Desktop 1.18.29 can offer OAuth models missing from active provider state[/+]oussama1598 commentedon Sep 5, 2026
Correction after tracing OpenCode 1.18.29 and testing locally:
ModelNotFoundError.messageintentionally renders${providerID}/${modelID}; the displayedopenai/gpt-...string was not evidence of a duplicated prefix.opencode models openai --refreshfollowed by a Desktop restart repaired the local state.openai/gpt-5.6-terrathen returned successfully.I updated the title and body so the report reflects the evidence. The remaining issue is stale provider/catalog state plus the silent Desktop failure, not model-ID parsing.
ianosei8 commentedon Sep 7, 2026
ty, i love you all 💕
thdxr commentedon Oct 11, 2026
Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (
opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.— from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊