Skip to content

Responses using GPT-5.3-Codex (from OpenAI Codex) terminate early #12570

@kanishkasahoo

Description

@kanishkasahoo

Description

When responding, the model makes one or two subagent calls, and stops before providing the next run. This behavior is absent from 5.2 on the same provider and task, and from other models like Opus 4.6

Plugins

None

OpenCode version

1.1.53

Steps to reproduce

  1. Select 5.3-Codex model from the picker (tested with OpenAI Codex as provider)
  2. Provide a prompt which causes multiple subagents to be invoked in sequence
  3. Observe that runs complete early (model ends its turn early before properly completinh task

Screenshot and/or share link

Image

Operating System

MacOS 26.2

Terminal

iTerm2

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions