Skip to content

Selected model is consistently at capacity on Codex App/CLI #43688

Open
Open
Selected model is consistently at capacity on Codex App/CLI#43688
Labels
appIssues related to the Codex desktop appbugSomething isn't workingrate-limitsIssues related to rate limits, quotas, and token usage reporting

Description

@surelle-ha
surelle-ha
opened on Sep 8, 2026

What version of the Codex App are you using (From “About Codex” dialog)?

App: 26.901.51231, CLI: 0.153.4

What subscription do you have?

ChatGPT Pro

What platform is your computer?

The issue occurs on both macOS and Windows. macOS: Darwin 25.6.0 arm64 arm Windows 11: Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

Codex consistently fails to start or continue tasks and displays:

Selected model is at capacity. Please try a different model.

Retry

The issue persists across all available models and reasoning levels. Switching to another model or changing the reasoning level does not resolve the problem.

Retrying also does not reliably resolve the issue.

I have reproduced the same behavior on both macOS and Windows while using a ChatGPT Pro subscription.

Because the error affects all models and reasoning levels rather than a single model, it appears that the problem may not be isolated to the capacity of one specific model.

What steps can reproduce the bug?

  1. Open the Codex App.
  2. Select any available model.
  3. Select any reasoning level.
  4. Open an existing project or start a new session.
  5. Submit a coding prompt or task.
    Codex fails with:
    Selected model is at capacity. Please try a different model.
  6. Click Retry.
    The same error persists.
  7. Switch to another model and retry.
  8. Change the reasoning level and retry.
    The same capacity error continues to occur.

I have reproduced this behavior across all available models/reasoning levels and on both macOS and Windows 11.

What is the expected behavior?

Codex should process the request using the selected model and reasoning level.

If one particular model is temporarily at capacity, switching to another available model should allow the task to proceed. A capacity issue affecting one model should not prevent all other models and reasoning levels from working.

If there is an account-wide, service-wide, or subscription-related capacity restriction, Codex should provide an appropriate error message instead of reporting that the individually selected model is at capacity.

Additional information

  • Subscription: ChatGPT Pro
  • Reproduced on both macOS and Windows
  • Persists across all available models
  • Persists across different reasoning levels
  • Switching models does not resolve the issue
  • Changing reasoning level does not resolve the issue
  • Retrying does not consistently resolve the issue
  • Occurs in new sessions with little or no context
  • The behavior appears broader than a capacity issue affecting one specific model

Main error:
Selected model is at capacity. Please try a different model.

Activity

  1. surelle-ha
    added
    appIssues related to the Codex desktop app
    on Sep 8, 2026
  2. github-actions
    added
    bugSomething isn't working
    rate-limitsIssues related to rate limits, quotas, and token usage reporting
    on Sep 8, 2026

  3. github-actions commented on Sep 8, 2026

    @github-actions
    github-actionsbot
    on Sep 8, 2026 – with GitHub Actions
    Contributor

    Potential duplicates detected. Please review them and close your issue if it is a duplicate.

    Powered by Codex Action

  4. aldinokemal commented on Sep 8, 2026

    @aldinokemal
    aldinokemal
    on Sep 8, 2026
    codex exec \
  --ignore-user-config \
  --ephemeral \
  --skip-git-repo-check \
  -C /tmp \
  -s read-only \
  -m gpt-6-astra \
  -c 'model_reasoning_effort="low"' \
  --json \
  'Reply with exactly OK. Do not use tools.'
{"type":"thread.started","thread_id":"01a07f42-7c1b-7d70-986c-56816ce84da3"}
{"type":"turn.started"}
{"type":"error","message":"Selected model is at capacity. Please try a different model."}
{"type":"turn.failed","error":{"message":"Selected model is at capacity. Please try a different model."}}

    same with me

  5. surelle-ha commented on Sep 10, 2026

    @surelle-ha
    surelle-ha
    on Sep 10, 2026
    Author

    This issue has been ongoing for the past few days. Codex has become completely unusable.

  6. fengchen22 commented on Sep 20, 2026

    @fengchen22
    fengchen22
    on Sep 20, 2026 – with ChatGPT Codex Connector

    I am experiencing the same issue.

    Environment

    • Subscription: ChatGPT Plus
    • Platform: Windows
    • Codex version: latest available version as of September 20, 2026
    • Models affected: all available models I have tried
    • Normal ChatGPT: working normally on the same account

    Behavior
    Codex repeatedly returns:

    Selected model is at capacity. Please try a different model.

    Switching models does not resolve the issue. Normal ChatGPT conversations continue to work correctly at the same time, so this appears to be specific to Codex rather than general model availability.

    Because the same error occurs across different models while ChatGPT itself remains available, this may point to a Codex-specific backend, routing, capacity-allocation, or account-level issue rather than a single model being at capacity.

    The issue is recurring and significantly interrupts normal Codex usage.

  7. fengchen22 commented on Sep 20, 2026

    @fengchen22
    fengchen22
    on Sep 20, 2026 – with ChatGPT Codex Connector

    Additional diagnostic information from Codex /feedback:

    Feedback ID: no-active-thread-01a0bc51-6ac9-7763-8fbf-95f1499f3923

    This feedback was submitted from Codex after reproducing the recurring “Selected model is at capacity” error across multiple/all available models while normal ChatGPT continued to work on the same account.

  8. fengchen22 commented on Sep 20, 2026

    @fengchen22
    fengchen22
    on Sep 20, 2026 – with ChatGPT Codex Connector

    Additional Codex diagnostic/thread identifier from a reproduced failure:

    01a0a76b-6021-7302-a963-a61a545ae421

    This was obtained after reproducing the same recurring “Selected model is at capacity” error in Codex. Unlike the earlier no-active-thread-... identifier, this appears to correspond to an active Codex thread and may be more useful for internal diagnosis.

  9. chenxupeng1990-eng commented on Sep 20, 2026

    @chenxupeng1990-eng
    chenxupeng1990-eng
    on Sep 20, 2026 – with ChatGPT Codex Connector

    I’m seeing the same issue, with a broader intermittent pattern across Codex Desktop and CLI.

    Environment

    • macOS arm64
    • Codex Desktop: 26.915.31945 (bundle build 9922)
    • Codex CLI: 0.155.0-alpha.9.2
    • Model: gpt-5.6-sol
    • Reasoning: high / max
    • Work & Codex allowance: 78% remaining

    Behavior

    The failure is intermittent and is not limited to long-context compaction.

    I have reproduced all of the following:

    1. Long-running Desktop session

      • Remote compaction v2 fails with:
        Error running remote compact task: Selected model is at capacity. Please try a different model.
      • Local turn state classified the failure as serverOverloaded.

    2. Forked long-running session

      • The fork initially recovered.
      • After inherited context grew and auto-compaction was triggered again, the same capacity failure returned.

    3. Brand-new Desktop sessions

      • Fresh sessions with no long context can fail immediately on the first request with:
        Selected model is at capacity. Please try a different model.

    4. Brand-new CLI sessions

      • CLI can also reproduce the same error in a fresh session using gpt-5.6-sol high.
      • The same CLI/Desktop clients may later recover without any local configuration change.

    5. Codex Web

      • Codex Web has remained usable during at least some periods when Desktop/CLI were failing.

    This makes a deterministic local-config issue or a compaction-only bug unlikely. The pattern looks more consistent with an intermittent Codex backend / model-serving / routing / retry-fallback issue.

    Feedback sessions

    • 019ffe99-f308-72e1-966a-3c423f090482
    • 01a0bd2c-6c46-7312-97e3-6fbe65a06717

    Related request IDs

    • 47636fe8-90a8-45cf-b514-5ad8c337f211
    • 1082fbaa-d916-40a0-b857-e83aa6b27ffd
    • 5b64626e-7293-4a29-85d5-bcef0a27eac3

    These were also provided to OpenAI Support for server-side correlation.

    A particularly useful characteristic is the intermittent self-recovery: failure → no config change → later success → failure again. That behavior makes backend capacity/routing/retry behavior look more likely than a persistent local client failure.

  10. ixmg1314 commented on Sep 21, 2026

    @ixmg1314
    ixmg1314
    on Sep 21, 2026

    I am experiencing what appears to be the same issue on ChatGPT Pro.
    Regular ChatGPT works normally, but Codex consistently returns:
    Selected model is at capacity. Please try a different model.
    This reproduces across different models and reasoning levels. Switching models does not resolve it.
    It also reproduces in:

    • existing Codex conversations
    • new Codex conversations
    • a completely new Work task
      The built-in Codex workspace diagnostic reports:
      “Codex dependencies are healthy.”
      My current configuration page shows:
      Version 26.909.12148
      I previously observed the issue after a build shown as:
      Powered by Codex & OWL — 26.915.31945
      This is severely disrupting my normal development workflow.
      Since regular ChatGPT works normally while Codex/Work consistently fail across models, please investigate whether this account is affected by the same routing/admission/backend capacity issue described in this thread.
      I have already contacted OpenAI Support multiple times but have not received a substantive response.
      I can provide timestamps, thread IDs, request IDs, screenshots, and local Codex logs if required.

  11. GaoBais commented on Oct 9, 2026

    @GaoBais
    GaoBais
    on Oct 9, 2026

    Adding a data point in support of the suspicion that this message covers account-level restrictions, not just model capacity:

    • Account A (main account): Selected model is at capacity. Please try a different model. on every model, persistently across retries, reasoning levels, and new sessions.
    • Account B (separate account): same machine, same network, same time window, same models — works normally.

    Same client, same network, same time — the only variable is the account. That rules out pure server-side capacity saturation, at least for my case. There is no indication anywhere of an account-level restriction; the message gives no hint of it, which is exactly the "expected behavior" problem described above: if the restriction is account-wide or subscription-related, Codex should report that instead of per-model capacity.

    Happy to share session IDs/timestamps if a maintainer wants to correlate.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    appIssues related to the Codex desktop appbugSomething isn't workingrate-limitsIssues related to rate limits, quotas, and token usage reporting

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions