---
title: GPT-5.6 Sol in Codex: keeping the work moving, a note by Arthur Galvão
url: https://arthurgalvao.work/notes/gpt-5-6-sol-in-codex
lang: en
---

# GPT-5.6 Sol in Codex: keeping the work moving, a note by Arthur Galvão

> Jul 16, 2026. Posted on [X](https://x.com/arthurgalvaorl/status/2077617721391399284), [LinkedIn](https://www.linkedin.com/feed/update/urn:li:activity:7483383175788982272/)

Markdown version of https://arthurgalvao.work/notes/gpt-5-6-sol-in-codex, the same content as the page, for AI agents and language models.

I’ve spent a good part of this week testing GPT-5.6 Sol inside OpenAI’s new ChatGPT experience, with Chat, Work and Codex now living in the same app.

What surprised me most was the depth of the model and how efficient it was with tokens.

I used it for some pretty heavy work in Codex, with the effort set to Ultra. We kept going for a long time before I finally hit the usage limit.

I’m on ChatGPT Plus, not Pro, so my included usage currently resets weekly. The additional resets OpenAI has made available have been a nice touch too. They gave me more room to keep working without upgrading or buying credits.

I still subscribe to Claude Max. After so much time using it, Claude already understands the way I build systems and projects really well. It’s still where I start most of this work.

But when my Claude usage runs out, Codex with GPT-5.6 Sol lets me pick up the work and keep building.

A few things I worked on with it this week:

A new version of the Archive page:
https://www.arthurgalvao.work/archive

A refreshed Work page:
https://www.arthurgalvao.work/work

And a new version of the About page:
https://www.arthurgalvao.work/about

Across both desktop and mobile, it handled this level of detail really well. That alone has made it a very useful part of my workflow.

[All notes](https://arthurgalvao.work/notes.md)