GPT-5.5 lands in ChatGPT and Codex as OpenAI shifts toward agent work

What the tweet revealed

OpenAI framed GPT-5.5 as a work model rather than a conversational novelty, writing A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion.. The main account then closed the loop with a deployment detail that matters more than the slogan: GPT-5.5 is already live in ChatGPT and Codex.

The OpenAI account is the company’s top-level channel for frontier model releases, and this post reads like a product line in the sand. It is not pitching a marginal chatbot upgrade. It is positioning GPT-5.5 as the model that should carry longer workflows across tools, files, and browser-like environments with less hand-holding from the user.

Context from the developer follow-ups

Two minutes later, OpenAI Developers added a more operational summary: GPT-5.5 is “our smartest frontier model yet” and is rolling out in ChatGPT and Codex today, with API access coming soon. A second follow-up broadened the Codex angle, saying browser use has expanded so Codex can test flows, click through web apps, capture screenshots, and iterate on what it sees until it completes the task. That matters because it turns the launch from a vague capability claim into a clear product direction: OpenAI wants GPT-5.5 to sit inside agent loops, not only in a chat box.

Because the direct OpenAI product page blocked clean public fetches during this crawl, the strongest public evidence available in the window is the coordinated tweet sequence itself plus the embed metadata. Even so, the picture is fairly crisp. The main account emphasizes complex goals, tool use, and self-checking; the developer account adds rollout timing and a concrete browser-use expansion for Codex.

What to watch next

The next signal is not marketing language but operating detail: API model IDs, pricing, system-card disclosures, and whether GPT-5.5 keeps its edge once third-party developers put it inside real QA, browser, and file-heavy workloads. If OpenAI follows the tweet sequence with stable API access and reproducible evaluation notes, GPT-5.5 will be judged less as a flagship chat model and more as the engine behind practical agent work.

Sources: X source tweet · OpenAI Developers rollout tweet · OpenAI Developers browser-use follow-up · Official embed