With GPT-5.5, Codex now gets more of the job done across the browser, files, docs, and your computer. We've expanded browser use so Codex can interact with web apps, and test flows, click through pages, capture screenshots, and iterate on what it sees until it completes the task.

Codex now generates higher-quality spreadsheets, slide decks, and documents with GPT-5.5 in @Microsoft Office and @googledrive . A new file viewer in the app makes it faster to review, revise, and iterate, so files are ready to share sooner.

Computer use just got better too. With GPT-5.5, Codex is stronger at using apps on your computer, from seeing what’s on screen to clicking, typing, navigating, and moving context across tools.

And in case you missed it, image generation in Codex also got better this week. With gpt-image-2, you can generate and edit visuals in the same workflow where you’re building apps, decks, mockups, and other artifacts.

Download the Codex app from

Claude is cooked..

Claude Code is 1000x's better and so much more reliable than this shifty, lying company which builds on quicksand. #opensource4o #return4o

When will this all come to the windows app?

"iterate on what it sees" is the key phrase. screenshot the result, identify the bug visually, fix it, repeat. that is the frontend testing loop devs actually need. Perplexity shipped something similar last week. OpenAI just brought it natively into the coding workflow.

Browser use is the real tell. Once the model can click, observe, retry, and continue until completion, the web stops being a reading surface and starts becoming a work surface

When is this coming for WIndows?

Codex must be the most unpopular product you released as opposed to 4o. Are you downgrading on purpose, or your employees are getting dumber? #bringback4o #keep4o

Two questions: 1) Available on windows and 2) for enterprise?

The interesting shift here isn’t just “new model,” it’s that OpenAI is explicitly framing GPT‑5.5 as infrastructure for agents and complex workflows, not chat. You can feel the product story moving from “AI that talks” to “AI that runs parts of your company’s operating system.”

very capable model. sadly doesn’t do quite well on SWE but crushes Mythos on Terminal Bench 2

Still Mac only I presume?

😒

Ok that makes shipping fast

Is this ever going to be available for windows?

Nope. not giving read/write access to spyware with ties to the military. Most immoral company. #QuitGPT

Why browser use over computer use ? I mean … is browser use actually better ?.. if I just computer use the same thing , is it … I mean I just can’t imagine browser use being something I utilized over computer use .. hopefully browser use is great enough to overtake cuse sometim

🚀 OpenAI just brought Browser Use into Codex. This is a big step for frontend development. Codex can now build the UI, open the app, click through it like a user, inspect console and network logs, catch bugs, and iterate on the fix. That closes the loop: build → test → debug → fix → verify This is where coding agents start becoming truly useful, not just impressive. Launched with GPT-5.5 in Codex. Who’s testing this first? 👀 #OpenAI #Codex #GPT55 #AI #DevTools #Frontend

This was very good, that's what I said.

It doesn’t work, GPT-5.5 in Codex {"type":"error","status":400,"error":{"type":"invalid_request_error","message":"The 'gpt-5.5' model is not supported when using Codex with a ChatGPT account."}}

Is browser use on windows yet?

Does codex have mobile control? I want to still get work done when I’m away from my laptop

@OpenAIDevs GPT-5.5 could control my phone as well

codex iterating on screenshots until a flow completes is exactly what QA automation tools charged $40k per year to do badly the first SaaS category this wipes out is no code testing platforms

#keep4o bring back 4o or open source 4o

This is where Codex feels really exciting to me. With GPT-5.5 improving browser use, files, docs, and computer use, it feels much closer to a real work agent. Being able to keep iterating until the task is done feels like a major step.

public agi is here, crazy

Better browser interaction was why Claude worked much better for me for development and debugging. If Codex improved that a lot, combined with images 2.0 this might make me switch towards Codex

Okay, so it’s finally time to move from Claude Code to Codex!

screenshots in the loop is the actual step forward. most agents break because they cant see what they just did

reset the limits , let us enjoying gpt 5.5!

Codex can click through pages now. your QA team has been warned

This is insane. Now codex is becoming the everything app for working