Repository navigation
[macOS][Desktop][Dot] Repeated “Selected model is at capacity” errors interrupt active work #52341
Description
Summary
While using Dot Lulu in an existing Codex conversation in the macOS desktop app, the selected model repeatedly stopped with this message:
Selected model is at capacity. Please try a different model.
This occurred several times in the same conversation during an ongoing coding task. The app displayed a retry countdown, but the capacity error returned again and interrupted progress.
Steps to reproduce
- Open the macOS desktop app.
- Open an existing project conversation and continue an ongoing task through Dot Lulu.
- Keep
GPT-6 Astra Ultraselected. - Let the Dot continue working, or retry after the capacity message appears.
Actual behavior
- The selected model becomes unavailable and shows the capacity message above.
- The same error recurs multiple times in the same conversation/session.
- Ongoing work is interrupted, and the user is told to choose a different model.
Expected behavior
Active Dot work should continue normally. A temporary capacity event should recover cleanly, or the app should provide a reliable fallback without repeatedly interrupting the same task.
Impact
Long-running coding work is repeatedly paused and requires user intervention or model switching.
Environment
- App: ChatGPT desktop, powered by Codex & OWL
- Version:
26.1002.52244 - Release date shown in app: October 7, 2026
- macOS: Golden Gate
27.0.1 - Device: 16-inch MacBook Pro, Apple M5 Max, 128 GB memory
- Dot: Lulu
- Conversation:
精简Along身份初始化 - Selected model:
GPT-6 Astra Ultra - Feedback ID:
01a11df0-448b-7548-96c8-a64c717ea0c0 - Occurrence: multiple times in the same conversation/session
An in-app feedback report was also submitted with ChatGPT diagnostic logs and browser logs enabled.
No other issue is being referenced here; this report is based on the directly observed repeated failure.
Activity
github-actions commentedon Oct 9, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
jhs1121 commentedon Oct 9, 2026
Additional report from a macOS Codex user (Oct 7–8, 2026, Pacific time):
Selected model is at capacity. Please try a different model.errors during active coding sessions over the last two days. The user has also reported encountering the same message after switching between model choices, rather than only one model.This report deliberately omits the user's project paths/files, account identifiers, and private code.
jhs1121 commentedon Oct 9, 2026
Additional diagnostic observation (follow-up to my earlier report):
During the Oct 7–8, 2026 period when Codex desktop repeatedly displayed
Selected model is at capacity. Please try a different model.across different model selections, ordinary ChatGPT web conversations remained usable for the same user. In contrast, Codex coding sessions were repeatedly interrupted by capacity errors, including after successful file edits and shell commands.This is a user-observed contrast, not a controlled simultaneous test of identical models or backend paths. It does not establish the root cause. However, it may help narrow the investigation toward Codex-specific model serving, agent continuation/scheduling, request routing, or account-level routing rather than assuming a general ChatGPT conversation outage.
Could the team check whether capacity errors can persist in Codex while ChatGPT chat requests succeed for the same account, and whether the UI is mapping other failures to this generic capacity message? No private account details or project content are included.