https://www.youtube.com/watch?v=6-I0-p-Lhug
This video announces and explains the new updates to OpenAI’s Codex, which now includes three new models:
- ChatGPT 5.1: This model is excellent for planning and generating creative ideas. The speaker emphasizes its effectiveness as a creative partner for brainstorming and structuring projects.
- ChatGPT 5.1 Codex: A version of 5.1 specifically optimized for larger, more complex coding tasks. It’s designed to be more thorough, using more tokens and taking more time to ensure the code is correct the first time.
- ChatGPT 5.1 Codex Mini: This is a faster, more lightweight model intended for smaller coding tasks like debugging, making styling changes, and other quick tweaks. It’s more efficient with tokens for these types of jobs.
The speaker demonstrates how to use these models within Visual Studio Code through the Codex extension. He outlines a new workflow:
- Use ChatGPT 5.1 in “chat mode” to plan out the application structure and features.
- Switch to “agent mode” and use ChatGPT 5.1 Codex for the initial, large-scale build of the application.
- For smaller fixes and refinements, use the faster and more efficient ChatGPT 5.1 Codex Mini.
The video also touches on the “Cloud Agents” feature in Codex, which allows users to run tasks in the cloud, even multiple versions of a task simultaneously, to compare results and select the best option. Finally, the speaker briefly compares Codex with Claude Code. While he notes that the new updates are a massive improvement for Codex, he personally still prefers the results from Claude Code (specifically Sonnet 4.5) for its design and aesthetic output, though he acknowledges that the gap between the two is closing.