exec not compatible with oss

Resolved 💬 2 comments Opened Jan 14, 2026 by blackbone Closed Jan 14, 2026

What version of Codex is running?

codex-cli 0.81.0

What subscription do you have?

plus but it doesn't matter

Which model were you using?

ollama gpt-oss:20b

What platform is your computer?

Darwin 24.6.0 arm64 arm

What issue are you seeing?

when using interactive mode with oss and local ollama everything seems works fine

% codex --oss --local-provider ollama "how r u?" 
╭────────────────────────────────────────────────╮
│ >_ OpenAI Codex (v0.81.0)                      │
│                                                │
│ model:     gpt-oss:20b high   /model to change │
│ directory: ~/git/project                       │
╰────────────────────────────────────────────────╯

but when executing with 'e' or 'exec' seems like oss flag is simply ignored

% codex --oss --local-provider ollama exec "how r u?"
OpenAI Codex v0.81.0 (research preview)
--------
workdir: /Users/blackbone/git/project
model: gpt-5.2-codex
provider: openai
approval: never
sandbox: workspace-write [workdir, /tmp, $TMPDIR]
reasoning effort: high
reasoning summaries: auto
session id: 019bbe60-36a7-7982-befa-eceeffe748ec
--------
user
how r u?
mcp: shadcn starting
mcp: atlassian starting
mcp: atlassian ready
mcp: shadcn ready
mcp startup: ready: atlassian, shadcn
codex
Doing well—thanks! What can I help with?
2026-01-14T21:18:59.505632Z ERROR codex_core::codex: needs_follow_up: false
tokens used
10,673
Doing well—thanks! What can I help with?

What steps can reproduce the bug?

  • run local ollama or use any oss setup which works in interactive mode
  • make non-interactive call with same oss model setup

What is the expected behavior?

cli uses oss model provider, model, etc

Additional information

_No response_

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗