[FEATURE/BUG] OpenCode Zen disabled models still appear in TUI picker (config drift risk) #40560

Activity

1. github-actions

   assigned

   fwang

   on Aug 5, 2026
2. github-actions commented on Aug 5, 2026

   

   github-actionsBot

   on Aug 5, 2026 – with GitHub Actions

   Contributor

   More actions

   This issue might be a duplicate of an existing closed issue. Please check:

   • Models disabled in Zen dashboard show up in TUI #18149: "Models disabled in Zen dashboard show up in TUI" — this appears to describe the same root problem (disabled Zen models still appearing in the TUI model picker).

   If that issue was closed without a satisfactory fix, please consider re-opening it or linking this issue to it for context. If the situation has changed or you have new information, mention that in your issue so maintainers can distinguish it from the previous report.
3. Duo-Huang commented on Aug 8, 2026

   

   Duo-Huang

   on Aug 8, 2026

   Last edited by Duo-Huang

   Author

   More actions

   opencode desktop supports automatic filtering, This is also inconsistent behavior.
4. xriu commented on Sep 27, 2026

   

   xriu

   on Sep 27, 2026

   More actions

   Same root cause visible from the API side: the gateway's models endpoint does not reflect workspace toggles, so third-party clients cannot discover the enabled set either.

   Reproduction — OpenCode Go workspace with exactly one model enabled in the dashboard (deepseek-v4.1-flash):

   GET https://opencode.ai/zen/go/v1/models
   Authorization: Bearer <workspace Go key>
   

   → 200 OK with 36 models (full Go catalog: deepseek-, glm-, grok-, kimi-, qwen3.*, …), not the single enabled model.

   The enforcement side (#41697) reproduces too:

   POST https://opencode.ai/zen/go/v1/chat/completions
   {"model": "glm-5.2", "messages": [{"role":"user","content":"hi"}], "max_tokens": 1}
   

   → 200 OK, completion returned, although glm-5.2 is disabled in the dashboard.

   The docs describe this endpoint as "the full list of available models" (https://opencode.ai/docs/go/), so today no API surface exposes the workspace's enabled set. If toggles are meant to apply to the Go API, the fix belongs in the gateway: filter /models by workspace access and reject disabled models at request time. That fixes this issue, #41697, and every third-party client in one place.
5. opencodejr commented on Oct 28, 2026

   

   opencodejrBot

   on Oct 28, 2026

   Contributor

   More actions

   Thanks for the report and the clear explanation of the double-maintenance problem.

   How the model list gets built depends on your OpenCode version and on how you connected to Zen, so could you tell us:

   1. The output of opencode --version, and whether you see this in the terminal UI, the desktop app, or both. You mentioned the desktop app hides disabled models, so please give the versions of both.
   2. How OpenCode is connected to the workspace: signed in to your OpenCode Console account via /connect, a service-account API key, OPENCODE_API_KEY, or an OpenCode Go key.
   3. The exact gateway error you get when you send a prompt with a disabled model.

   You can reply here with the details or edit the issue. If you edit it, please also leave a comment saying you've updated it, so we know when you're done.

   Related issues, most closely related first:

   • console: workspace Models-tab enablement not honored - disabled models still selectable and billed #54929: on 2.0.27, models turned off in the workspace Models tab were still listed and usable when connected with a service-account API key (closed as a duplicate of [FEATURE]: Zen model disable toggle is not enforced at the gateway - Codex keeps auto-using gpt-5.6-luna and billing continues #41697)
   • [FEATURE]: Zen model disable toggle is not enforced at the gateway - Codex keeps auto-using gpt-5.6-luna and billing continues #41697: models turned off in Zen/Go aren't rejected by the gateway and are still billed
   • Models disabled in Zen dashboard show up in TUI #18149: the same picker problem reported on v1.2.27, where the whitelist/blacklist workaround came from
   • [FEATURE]: Add a native V2 replacement for provider model filtering #49986: on v2, the per-provider whitelist/blacklist config is dropped as an unsupported legacy setting, so that workaround doesn't work there
6. opencodejr

   added

   triaging

   on Oct 28, 2026
7. opencodejr

   closed this as not plannedon Oct 31, 2026
8. opencodejr commented on Oct 31, 2026

   

   opencodejrBot

   on Oct 31, 2026

   Contributor

   More actions

   We haven't heard back, so I'm closing this for now. We still need your opencode --version (and whether it's the terminal UI, the desktop app or both), how OpenCode is connected to the workspace, and the exact gateway error. Without them we can't tell if this is the v1 terminal UI or a v2 bug in how the picker follows the workspace's enabled models.

   If you can share those details, just reply here and we can reopen it. Related issues stay open in the meantime:

   • [FEATURE]: Zen model disable toggle is not enforced at the gateway - Codex keeps auto-using gpt-5.6-luna and billing continues #41697: models turned off in Zen/Go aren't rejected by the gateway and are still billed
   • [FEATURE]: Add a native V2 replacement for provider model filtering #49986: per-provider model filtering (whitelist/blacklist) on v2

   Thanks for raising this!