The new Codex for developers

YouTube AI Channels Products

Summary

OpenAI merges Codex with ChatGPT, introducing GPT-5.6 Sol, Ultra mode, sub-agents, Sites one-click deployment, task management, mobile support, and inline PR review for developers.

No content available
Original Article
View Cached Full Text

Cached at: 07/15/26, 03:17 AM

TL;DR: Codex merges with ChatGPT, introducing GPT-5.6 Sol model, Ultra mode, sub-agents, appshot screenshots, Sites one-click deployment, task management, mobile support, and inline PR review. ## The New Codex: Merged with ChatGPT Today, over 6 million users use Codex every week. Based on your feedback, Codex and ChatGPT are now merged into one app. For developers, it's still the familiar Codex, but now with a dedicated space in ChatGPT, right next to the new work agents. Over the past two months, they've shipped more than 150 updates, and this is just the beginning. ## Models and Ultra Mode ### GPT-5.6 Sol GPT-5.6 Sol is the latest frontier model, available to everyone. It can handle complex problems over long periods, and you can use the new **Ultra mode** directly on GPT-5.6. ### Ultra Mode + Goal Instructions With Goal instructions, you can give it your most challenging projects. When using Ultra mode, Sol gets a larger reasoning budget. But be aware, Ultra mode will consume your token limit faster. ### Sub-agents Working in Parallel Codex automatically breaks work into multiple sub-agents without needing manual requests. You can track any individual sub-agent or open a panel to see the entire team working together. One goal, all work progressing in parallel. ## Computer and Browser Operations Sol excels at computer and browser operations. It's even faster with GPT-5.6. For example, when developing an iPhone app, pressing Cmd-Cmd on a Mac captures an **appshot** of the simulator (a smarter screenshot that also grabs screen and app context). Then say: "Navigate to each screen and capture App Store screenshots in both English and French." Codex directly operates the simulator — it's not your cursor, Codex is clicking, and it does all this in the background while you continue using your computer. ### Modify Apps in the Browser You can open a simple paper airplane game in the in-app browser, play it, and see exactly what Codex created. To modify something, click to annotate, and Codex gets the visual context and makes the change. The browser now includes support for apps that require login and passkeys. After making changes, you can view a diff view or even edit inline — you can directly modify a line of code or describe the desired behavior and let Codex work. It feels more like true pair programming with the model. ## Sites: One-Click Deployment **Sites**, launched last month for Enterprise, is now available to all users and works with any web app you build with Codex. Just say "Publish this to Sites," and it comes with built-in hosting, authentication, persistent database, and file storage. You can go from an idea to a real full-stack app without worrying about infrastructure or hosting. ## Task Management Codex can also help manage its own Codex tasks. For example, find five bugs from Linear, create a separate task and an isolated work tree for each one, and mark the most critical ones. From the coordination thread, you can see all tasks started and progressing. Codex isn't just doing the work; it's also helping you stay organized and connected. Tasks can also reference each other. If an idea originated in ChatGPT or used Deep Research, you can now @mention that conversation to bring the context directly into Codex — no more copy-pasting. ## Mobile Features ChatGPT on mobile gets a massive upgrade: 1. Create, search, open, and manage tasks directly on your phone. 2. Review changes: filter staged, unstaged, branches, or compare branches on mobile devices. 3. Directly connect to SSH hosts (e.g., dev machines) and start new tasks from your phone. You can start something at your desk, continue on your phone, and step in whenever Codex needs you. ## Pull Request Handling Codex makes it easier to handle pull requests without leaving the app. It knows which branch each task works on, and when that branch has an open PR, the summary panel automatically shows it. You can track everything, address issues, and merge when checks pass. If a check fails or you receive comments, have Codex fix them with full context. The new "Pull Requests" tab lets you explore each code change, leave inline comments, complete a full review, and mark your own PRs as ready and merge. All of this happens inside Codex — no need to go to GitHub.com, keeping you in flow. ## Closing and Build Week This is just a small part of what's been released, but the most important thing remains unchanged: continuously improving Codex for developers. We hope you'll join OpenAI Build Week, either online or at one of nearly 30 events around the world. > Alright, Codex, my demo is done. Can you help developers get ready for Build Week and then wrap up? As always, keep sending us your trickiest bugs, your boldest ideas, and those jobs that seem a bit impossible. Thanks. Source: The new Codex for developers - OpenAI (https://www.youtube.com/watch?v=eiQgljOrkWU)

Similar Articles

Introducing upgrades to Codex

OpenAI Blog

OpenAI releases GPT-5-Codex, an optimized version of GPT-5 specialized for agentic software engineering tasks, available via API and across Codex's integrated development environment with improved code review capabilities and long-form task execution.

Introducing Codex

OpenAI Blog

OpenAI launches Codex, a cloud-based AI software engineering agent powered by codex-1 (optimized o3), capable of writing features, fixing bugs, and proposing pull requests with parallel task execution. Available to ChatGPT Pro, Business, Enterprise users today, with Plus and Edu support coming soon.

Introducing GPT-5.2-Codex

OpenAI Blog

OpenAI releases GPT-5.2-Codex, an advanced agentic coding model optimized for complex software engineering tasks with improvements in long-context understanding, Windows support, and cybersecurity capabilities. The model achieves state-of-the-art performance on SWE-Bench Pro and Terminal-Bench 2.0, and is now available to paid ChatGPT users with API access coming in the following weeks.

Addendum to GPT-5 system card: GPT-5-Codex

OpenAI Blog

OpenAI has released GPT-5-Codex, a version of GPT-5 optimized for agentic coding tasks, trained with reinforcement learning on real-world coding environments. It is available via Codex CLI, IDE extensions, GitHub, and ChatGPT mobile, with comprehensive safety measures including sandboxing and prompt injection mitigations.