> ## Content Index
> Fetch the complete content index at: https://localmodelbench.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# OpenAI GPT-5.5 (Codex CLI): what clean workflow closure looks like
- URL: https://localmodelbench.com/codex-reference/
- Published: 2026-05-16T09:00:00.000Z
- Updated: 2026-09-11T16:45:07.000Z
- Description: Codex is not a local LM Studio run. It is kept as a reference line for what stronger agentic tooling does on the same public cases, using the same artifacts and hidden-oracle checks.
- Author: Local Model Bench Admin
- Tags: reference, note

The Codex row is intentionally not presented as a local-model victory lap. It is a ceiling/reference line: a stronger agentic coding environment running the same public paperwork cases.

That reference is useful because it shows which failures are benchmark difficulty rather than local-only weakness. Even the reference run still missed proof/evidence details, which is exactly why Local Model Bench separates core pass from resolved pass.

## Where it worked

- Best current practical score across the full public case set.
- Strong at preserving protected input folders while producing required artifacts.
- Most failures were narrow near misses rather than broad document misunderstanding.

## Where it failed

- Still failed one case strictly and had proof/evidence misses.
- Not directly comparable to local-only LM Studio runs.
- Useful as a ceiling/reference, not as the point of the site.