Hi, this is my first post in the community.
Since July 17, 2026, Codex Desktop has repeatedly stopped active coding tasks with the following message:
Selected model is at capacity. Please try a different model.
Environment
- Product: Codex Desktop for Windows
- Selected model: GPT-5.6
- Reasoning level: High
- Task type: Local repository work with full access
- Operating system: Windows
- Approximate time: July 17, 2026, during the afternoon in Spain (CEST)
What happens
The agent starts normally and successfully executes several local commands. During the same task, it suddenly stops and displays the capacity error.
After submitting a follow-up message, it may continue for a short time, execute more commands, and then show the same error again.
This has happened multiple times in the same session and interrupts long-running repository tasks before Codex can finish the requested validation, commit, and push.
What I have tried
- Retrying the same task
- Sending a follow-up message to continue
- Waiting and trying again
- Checking the OpenAI status page
- Confirming that the application still has access to the local repository
Questions
- Is this a temporary capacity limitation affecting GPT-5.6?
- Is there a way to automatically fall back to another model without losing the task context?
- Could this be related to an account-level usage or rate limit even though the message specifically says that the model is at capacity?
- What diagnostic information should I collect if this continues?
I can provide the Codex app version, timestamps, sanitized screenshots, and additional logs if needed.
Thank you.
