Clarification on Codex CLI models vs ChatGPT UI models

[onegigbyte]

Hi team, thanks for the great work on Codex CLI — it’s been very useful so far. I have a few questions that I think would help users understand the differences between CLI and the web UI.

Codex CLI has /model: minimal, low, medium, high.
ChatGPT UI has models: Instant, Thinking, Pro.

It’s not clear how these sets line up, and I’d love clarification on a few points:

1. Mapping: How do the CLI models map to the models on web UI?
2. Models: Are the underlying models the same, or are there differences specific to Codex CLI? Any benchmarks or details would help.
3. Verbosity: Codex CLI outputs are consistently much shorter than ChatGPT UI responses. Is there a way to enable richer outputs (similar to the UI’s style)?
4. Future consistency: Is it feasible or on the roadmap for Codex CLI to eventually return the same results as the web UI (other than search and deep research), so behavior is consistent across tools?

Having clarity on these would make it easier to choose the right setting for different tasks and know what trade-offs to expect.

Thanks for considering!