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
{{ message }}
Repository navigation
desktop: gpt-6.1-sol missing from ChatGPT OAuth models, shared session model forcibly switched #52375
Same ChatGPT Plus account via browser OAuth: CLI v2.0.10 lists and uses openai/gpt-6.1-sol fine, but Desktop 2.0.20 doesn't list it under OpenAI. Opening the shared CLI session in Desktop then forcibly switches its model to a Desktop-available one.
Active plugins: none (global config has no plugins key)
Reproduction
Connect OpenAI with ChatGPT browser OAuth (Plus account).
CLI v2.0.10: select openai/gpt-6.1-sol, send a message → responds normally.
Desktop 2.0.20: open /models → gpt-6.1-sol absent from the OpenAI section (only up to GPT-5.6-Sol / GPT-6 Astra family). The Console provider DOES list GPT-6.1 Sol.
With the CLI session open, launch Desktop → the session model is switched to a Desktop-available model (Muse Spark 1.3 Free via OpenCode provider in my case).
Expected Behavior
gpt-6.1-sol listed and selectable under OpenAI OAuth in Desktop (as in CLI, Console, official Codex, and per OpenAI: Plus/Pro in Codex), and shared sessions keep their model.
Actual Behavior
Model missing in Desktop; shared session's model forcibly replaced when Desktop is launched.
Additional Context
Suspected: the token-sharing plugin (packages/core/src/plugin/provider/chatgpt.ts, added in feat(core): add separate ChatGPT token-sharing OAuth method #52147, present in 2.0.20 but not in CLI 2.0.10) rebuilds the OpenAI model list from GET api.openai.com/v1/models filtered by visibility === "list" && supported_in_api and disables the rest; its fallbackModels also lacks gpt-6.1-sol. My Desktop list includes Astra Ultrafast (not in the fallback set), so the remote fetch succeeded but the slug is still absent.
CLI v2.0.10 uses the older Codex-backend path (provider/openai.ts, major > 5 passes 6.1) → works.
Tried a providers.openai.models override in global config → no effect (reverted), consistent with the list being rebuilt from the remote response.
Another report shows the same failure mode with openai/gpt-6-luna: it was previously selected and working, then disappeared from /models, and the active session switched to Muse Spark 1.3 Free without an explicit user choice.
The affected model ID differs from the one in this issue (gpt-6.1-sol), but the user-visible behavior is the same. The report also shows OpenCode retrying with The backend is temporarily overloaded before the availability/switch problem.
Please check whether model-catalog refresh or fallback handling can silently replace the selected model. The expected behavior should be to preserve the selected model, show a clear error, and require explicit confirmation before routing a session to another provider or model.
Verified fixed on my side with the latest Desktop update.
Environment: Desktop 2.0.24 / Windows 11 / ChatGPT Plus (browser OAuth, same account as before)
Verified:
openai/gpt-6.1-sol is now listed under OpenAI and responds normally.
openai/gpt-6-sol and openai/gpt-6-luna are also listed now (both were missing before, same as in my earlier check).
Opening a CLI session that uses openai/gpt-6.1-sol in Desktop no longer switches the model — the session keeps gpt-6.1-sol.
So all three symptoms from this issue are resolved in my environment. Leaving the issue open for the other reporters (gpt-6-luna forced-switch report, CLI 2.0.21 report) and for the maintainers to close.
Thanks for confirming that the models are available again and opening the shared session in Desktop no longer changes its model. I'm closing this as resolved; if it comes back on the latest v2, please open a new issue with your version and steps.
Activity
github-actions commentedon Sep 30, 2026
This issue might be a duplicate of existing issues. Please check:
mynameistito commentedon Oct 5, 2026
Additional report
Another report shows the same failure mode with
openai/gpt-6-luna: it was previously selected and working, then disappeared from/models, and the active session switched toMuse Spark 1.3 Freewithout an explicit user choice.The affected model ID differs from the one in this issue (
gpt-6.1-sol), but the user-visible behavior is the same. The report also shows OpenCode retrying withThe backend is temporarily overloadedbefore the availability/switch problem.Please check whether model-catalog refresh or fallback handling can silently replace the selected model. The expected behavior should be to preserve the selected model, show a clear error, and require explicit confirmation before routing a session to another provider or model.
drhanfastolfe commentedon Oct 5, 2026
I'm also affected on OpenCode CLI 2.0.21: openai/gpt-6.1-sol is missing from the OpenAI model list despite a successful OAuth connection.
This affects the CLI too, not just Desktop, and prevents me from selecting this model in OpenCode.
Is there a confirmed workaround for current v2 releases without downgrading, or any update on a fix? Thanks.
SDSDFFG-rgb commentedon Oct 7, 2026
Verified fixed on my side with the latest Desktop update.
Environment: Desktop
2.0.24/ Windows 11 / ChatGPT Plus (browser OAuth, same account as before)Verified:
openai/gpt-6.1-solis now listed under OpenAI and responds normally.openai/gpt-6-solandopenai/gpt-6-lunaare also listed now (both were missing before, same as in my earlier check).openai/gpt-6.1-solin Desktop no longer switches the model — the session keepsgpt-6.1-sol.So all three symptoms from this issue are resolved in my environment. Leaving the issue open for the other reporters (
gpt-6-lunaforced-switch report, CLI 2.0.21 report) and for the maintainers to close.neriousy commentedon Oct 20, 2026
Thanks for confirming that the models are available again and opening the shared session in Desktop no longer changes its model. I'm closing this as resolved; if it comes back on the latest v2, please open a new issue with your version and steps.