BuildBrief: (NEW) Claude 4.5 vs GPT-5 Codex

Anthropic released Claude 4.5 this week, and I've been testing it against GPT-5 Codex for the past few days whilst building.

Here's what actually matters:

Claude 4.5 is fast... very fast.

It jumps into code quickly, which feels productive at first...

But it often skips steps, rushes through planning, and writes tons of unnecessary code.

Claude 4.5 is definitely better than the previous versions, but it still writes so much cluttered code.

GPT-5 Codex is slower, but it thinks better and plans more carefully

-- and the quality of the output is much better.

 

When I use Codex:

1. Building new features
2. Cleaning up the codebase
3. When I need to change more than just 1 specific file

 

When I use Claude 4.5:

1. Quick, targeted changes (ex: UI fix, color change, etc)

 

I have Codex as an extension in Cursor (only need to pay for ChatGPT Plus).

And it doesn't wreck the entire codebase with useless code.

Here's the link to add the extension: Click here!

Happy building!
– Kristoffer, BuildWithAI