Latest updates to Codex, OpenAI’s coding agent
With the new extension for Chrome, Codex is even better at working with apps and websites in your browser. It works in parallel across tabs in the background without taking over your browser, and you stay in control of which websites Codex can use.
Learn more in the Codex Chrome extension documentation.
Updated the Codex enterprise governance guide with more detailed coverage of the Analytics dashboard charts, data export options, and enterprise Analytics API endpoints.
GPT-5.5 is now available in Codex as OpenAI’s newest frontier model for complex coding, computer use, knowledge work, and research workflows.
GPT-5.5 is the recommended choice for most Codex tasks when it appears in your model picker. It’s especially useful for implementation, refactors, debugging, testing, validation, and knowledge-work artifacts.
To switch to GPT-5.5:
codex --model gpt-5.5/model during a session.If you don’t see GPT-5.5 yet, update the CLI, IDE extension, or Codex app to the latest version. During the rollout, continue using GPT-5.4 if GPT-5.5 is not yet available.
The Codex app can now let Codex operate the in-app browser for local development servers and file-backed pages. Ask Codex to use the browser when it needs to click through a rendered UI, reproduce a visual bug, or verify a local fix inside the app.
Browser use runs through the bundled Browser plugin. In settings, you can manage the plugin and review allowed or blocked websites.
Codex can route eligible approval prompts through an automatic reviewer agent before the request runs. When configured, the Codex app shows an automatic review item with the review status and risk level, so you can see whether the reviewer approved, denied, stopped, or timed out before deciding.
Codex is becoming a broader workspace for getting work done with AI. This update makes it easier to start work with less setup, verify what Codex is building, create richer outputs, and keep momentum across longer-running tasks.
The Codex app now includes an early in-app browser. You can open local or public pages that don’t require sign-in, comment directly on the rendered page, and ask Codex to address page-level feedback.
Computer use lets Codex operate macOS apps by seeing, clicking, and typing, which helps with native app testing, simulator flows, low-risk app settings, and GUI-only bugs.
The feature isn’t available in the European Economic Area, the United Kingdom, or Switzerland at launch.
Chats are threads you can start without choosing a project folder first. They’re useful for research, writing, planning, analysis, source gathering, and tool-driven work that doesn’t begin in a codebase.
For work that needs a later check-in, thread automations can wake up the same thread on a schedule while preserving the conversation context. Use them to check a long-running process, watch for updates, or continue a follow-up loop without starting from scratch.
The task sidebar makes plans, sources, generated artifacts, and summaries easier to follow while Codex works. Context-aware suggestions can also help you pick up relevant follow-ups when you start or return to Codex.
Codex now brings more of the pull request workflow into the app. You can inspect GitHub pull requests in the sidebar, review comments in the diff, review changed files, then ask Codex to explain feedback, make changes, check them, and keep the review moving.
The artifact viewer can preview generated files such as PDF files, spreadsheets, documents, and presentations in the sidebar before you commit or share them. Memories, where available, can also carry useful context from past tasks into future threads, including stable preferences, project conventions, and recurring work patterns.
We’re updating model availability for users who sign in with ChatGPT. Starting
April 7, the model picker no longer shows gpt-5.2-codex,
gpt-5.1-codex-mini, gpt-5.1-codex-max, gpt-5.1-codex, gpt-5.1, or
gpt-5. On April 14, we’ll remove those models from Codex for ChatGPT sign-in.
Users can still choose from gpt-5.4, gpt-5.4-mini, gpt-5.3-codex, and
gpt-5.2. ChatGPT Pro users can also choose gpt-5.3-codex-spark.
To use another API-supported model in Codex, sign in with an API key or configure a model provider.
Codex now supports plugins: installable bundles that package skills, app integrations, and MCP server configuration for reusable workflows.
Plugins are available in the Codex app, CLI, and IDE extensions.
You can install curated plugins from the plugin directory, or scaffold a local
plugin with @plugin-creator and test it with workspace-scoped or home-scoped
marketplaces.
Learn more in the plugins documentation.
Every plugin is a folder with a required .codex-plugin/plugin.json manifest
and optional supporting files:
my-plugin/
.codex-plugin/
plugin.json # Required: plugin manifest
skills/ # Optional: packaged skills
.app.json # Optional: app or connector mappings
.mcp.json # Optional: MCP server configuration
assets/ # Optional: icons, logos, screenshotsYou can install plugins for just yourself with
~/.agents/plugins/marketplace.json and ~/.codex/plugins/, or for everyone
on a project with .agents/plugins/marketplace.json and a repo-local plugin
directory such as ./plugins/.
Codex surfaces curated public plugins in the plugin directory. Codex also ships
with the built-in @plugin-creator skill to help you scaffold a plugin, add a
local marketplace entry, and test it before sharing it with teammates.
@ menu so you can insert them from the composer alongside other mentions.Cmd/Ctrl+F now starts with your current text selection, which makes searching reviews and diffs faster.GPT-5.4 mini is now available in Codex as a fast, efficient model for lighter coding tasks and subagents.
It improves over GPT-5 mini across coding, reasoning, image understanding, and tool use while running more than 2x faster. In Codex, GPT-5.4 mini uses 30% as much of your included limits as GPT-5.4, so comparable tasks can last about 3.3x longer before you hit those limits.
GPT-5.4 mini is available in the Codex app, the CLI, the IDE extension, and Codex on the web. GPT-5.4 mini is also available in the API.
Use GPT-5.4 mini for codebase exploration, large-file review, processing supporting documents, and other less reasoning-intensive subagent work. For more complex planning, coordination, and final judgment, start with GPT-5.4.
To switch to GPT-5.4 mini:
codex --model gpt-5.4-mini/model during a session.If you don’t see GPT-5.4 mini yet, update the CLI, IDE extension, or Codex app to the latest version.
Change the Codex app appearance in Settings by choosing a base theme, adjusting accent, background, and foreground colors, and changing the UI and code fonts. You can also share your custom theme with friends.
You can now choose whether automations run locally or on a worktree, define custom reasoning levels and models, and use templates to find inspiration for new automations.
Various bug fixes and performance improvements.
GPT-5.4 is now available in Codex as OpenAI’s most capable and efficient frontier model for professional work.
It combines recent advances in reasoning, coding, and agentic workflows in one model, and it’s the recommended choice for most Codex tasks.
In Codex, GPT-5.4 is the first general-purpose model with native computer-use capabilities. GPT-5.4 in Codex includes experimental support for the 1M context window. It supports complex workflows across applications and long-horizon tasks, with stronger tool use and tool search that help agents find and use the right tools more efficiently.
GPT-5.4 is available everywhere you can use Codex: the Codex app, the CLI, the IDE extension, and Codex Cloud on the web. GPT-5.4 is also available in the API.
To switch to GPT-5.4:
codex --model gpt-5.4/model during a session.If you don’t see GPT-5.4 yet, update the CLI, IDE extension, or Codex app to the latest version.
codex://new links could stop prefilling a new conversation when the app was already open.The Codex app is now available on Windows. The app gives you one interface for working across projects, running parallel agent threads, and reviewing results in one place.
The Codex app runs natively on Windows using PowerShell and a native Windows sandbox for bounded permissions, so you can use Codex on Windows without moving your workflow into WSL, onto a virtual machine, or by deactivating the sandbox.
The Windows app includes the same core features as the rest of the Codex app:
If you prefer to develop in WSL, you can also switch the Codex agent and the integrated terminal to run there.
Download it from the Microsoft Store and sign in with your ChatGPT account or an API key. For setup and configuration details, see Setup, Use WSL with the Codex app, and Customize the app for your development setup.
@mentions and skill mentions in inline review comments.Today, we’re releasing a research preview of GPT-5.3-Codex-Spark, a smaller version of GPT-5.3-Codex and our first model designed for real-time coding. Codex-Spark is optimized to feel near-instant, delivering more than 1000 tokens per second while remaining highly capable for real-world coding tasks.
Codex-Spark is available in research preview for ChatGPT Pro users in the latest Codex app, CLI, and IDE extension. This release also marks the first milestone in our partnership with Cerebras.
At launch, Codex-Spark is text-only with a 128k context window. During the research preview, usage has separate model-specific limits and doesn’t count against standard Codex limits. During high demand, access may slow down or queue while we balance reliability across users.
To switch to GPT-5.3-Codex-Spark:
codex --model gpt-5.3-codex-spark/model during a session.If you don’t see GPT-5.3-Codex-Spark yet, update the CLI, IDE extension, or Codex app to the latest version.
GPT-5.3-Codex-Spark isn’t available in the API at launch.
For API-key workflows, continue using gpt-5.2-codex.
Alpha testing for the Codex app on Windows is also starting. Sign up here to be a potential alpha tester.
plan in the composer.Starting today, GPT-5.3-Codex is available natively in Cursor and VS Code.
API access is starting with a small set of customers as part of a phased release.
This is the first model treated as a high security capability under the Preparedness Framework.
Safety controls will continue to scale, and API access will expand over the next few weeks.
Today we’re releasing GPT-5.3-Codex, the most capable agentic coding model to date for complex, real-world software engineering.
GPT-5.3-Codex combines the frontier coding performance of GPT-5.2-Codex with stronger reasoning and professional knowledge capabilities, and runs 25% faster for Codex users. It’s also better at collaboration while the agent is working—delivering more frequent progress updates and responding to steering in real time.
GPT-5.3-Codex is available with paid ChatGPT plans everywhere you can use Codex: the Codex app, the CLI, the IDE extension, and Codex Cloud on the web. API access for the model will come soon.
To switch to GPT-5.3-Codex:
codex --model gpt-5.3-codex/model during a session.For API-key workflows, continue using gpt-5.2-codex while API support rolls
out.
The Codex app for macOS is a desktop interface for running agent threads in parallel and collaborating with agents on long-running tasks. It includes a project sidebar, thread list, and review pane for tracking work across projects.
Key features:
For a limited time, ChatGPT Free and Go include Codex, and Plus, Pro, Business, Enterprise, and Edu plans get double rate limits. Those higher limits apply in the app, the CLI, your IDE, and the cloud.
Learn more in the Introducing the Codex app blog post.
Check out the Codex app documentation for more.
Codex now enables web search for local tasks in the Codex CLI and IDE Extension.
By default, Codex uses a web search cache, which is an OpenAI-maintained index of web results. Cached mode returns pre-indexed results instead of fetching live pages, while live mode fetches the most recent data from the web. If you are using --yolo or another full access sandbox setting, web search defaults to live results. To disable this behavior or switch modes, use the web_search configuration option:
web_search = "cached" (default; serves results from the web search cache)web_search = "live" (fetches the most recent data from the web; same as --search)web_search = "disabled" to remove the toolTo learn more, check out the configuration documentation.
Team Config groups the files teams use to standardize Codex across repositories and machines. Use it to share:
config.toml defaultsrules/ for command controls outside the sandboxskills/ for reusable workflowsCodex loads these layers from .codex/ folders in the current working directory, parent folders, and the repo root, plus user (~/.codex/) and system (/etc/codex/) locations. Higher-precedence locations override lower-precedence ones.
Admins can still enforce constraints with requirements.toml, which overrides defaults regardless of location.
Learn more in Team Config.
Custom prompts are now deprecated. Use skills for reusable instructions and workflows instead.
GPT-5.2-Codex is now available in the API and for users who sign into Codex with the API.
To learn more about using GPT-5.2-Codex check out our API documentation.
Codex now supports agent skills: reusable bundles of instructions (plus optional scripts and resources) that help Codex reliably complete specific tasks.
Skills are available in both the Codex CLI and IDE extensions.
You can invoke a skill explicitly by typing $skill-name (for example, $skill-installer or the experimental $create-plan skill after installing it), or let Codex select a skill automatically based on your prompt.
Learn more in the skills documentation.
Following the open agent skills specification, a skill is a folder with a required SKILL.md and optional supporting files:
my-skill/
SKILL.md # Required: instructions + metadata
scripts/ # Optional: executable code
references/ # Optional: documentation
assets/ # Optional: templates, resourcesYou can install skills for just yourself in ~/.codex/skills, or for everyone on a project by checking them into .codex/skills in the repository.
Codex also ships with a few built-in system skills to get started, including $skill-creator and $skill-installer. The $create-plan skill is experimental and needs to be installed (for example: $skill-installer install the create-plan skill from the .experimental folder).
Codex ships with a small curated set of skills inspired by popular workflows at OpenAI. Install them with $skill-installer, and expect more over time.
Today we are releasing GPT-5.2-Codex, the most advanced agentic coding model yet for complex, real-world software engineering.
GPT-5.2-Codex is a version of GPT-5.2 further optimized for agentic coding in Codex, including improvements on long-horizon work through context compaction, stronger performance on large code changes like refactors and migrations, improved performance in Windows environments, and significantly stronger cybersecurity capabilities.
Starting today, the CLI and IDE Extension will default to gpt-5.2-codex for users who are signed in with ChatGPT. API access for the model will come soon.
If you have a model specified in your config.toml configuration file, you can instead try out gpt-5.2-codex for a new Codex CLI session using:
codex --model gpt-5.2-codexYou can also use the /model slash command in the CLI. In the Codex IDE Extension you can select GPT-5.2-Codex from the dropdown menu.
If you want to switch for all sessions, you can change your default model to gpt-5.2-codex by updating your config.toml configuration file:
model = "gpt-5.2-codex”Assign or mention @Codex in an issue to kick-off a Codex cloud task. As Codex works, it posts updates back to Linear, providing a link to the completed task so you can review, open a PR, or keep working.
To learn more about how to connect Codex to Linear both locally through MCP and through the new integration, check out the Codex for Linear documentation.
Minor updates to address a few issues with Codex usage and credits:
Today we are releasing GPT-5.1-Codex-Max, our new frontier agentic coding model.
GPT‑5.1-Codex-Max is built on an update to our foundational reasoning model, which is trained on agentic tasks across software engineering, math, research, and more. GPT‑5.1-Codex-Max is faster, more intelligent, and more token-efficient at every stage of the development cycle–and a new step towards becoming a reliable coding partner.
Starting today, the CLI and IDE Extension will default to gpt-5.1-codex-max for users that are signed in with ChatGPT. API access for the model will come soon.
For non-latency-sensitive tasks, we’ve also added a new Extra High (xhigh) reasoning effort, which lets the model think for an even longer period of time for a better answer. We still recommend medium as your daily driver for most tasks.
If you have a model specified in your config.toml configuration file, you can instead try out gpt-5.1-codex-max for a new Codex CLI session using:
codex --model gpt-5.1-codex-maxYou can also use the /model slash command in the CLI. In the Codex IDE Extension you can select GPT-5.1-Codex from the dropdown menu.
If you want to switch for all sessions, you can change your default model to gpt-5.1-codex-max by updating your config.toml configuration file:
model = "gpt-5.1-codex-max”Along with the GPT-5.1 launch in the API, we are introducing new gpt-5.1-codex-mini and gpt-5.1-codex model options in Codex, a version of GPT-5.1 optimized for long-running, agentic coding tasks and use in coding agent harnesses in Codex or Codex-like harnesses.
Starting today, the CLI and IDE Extension will default to gpt-5.1-codex on macOS and Linux and gpt-5.1 on Windows.
If you have a model specified in your config.toml configuration file, you can instead try out gpt-5.1-codex for a new Codex CLI session using:
codex --model gpt-5.1-codexYou can also use the /model slash command in the CLI. In the Codex IDE Extension you can select GPT-5.1-Codex from the dropdown menu.
If you want to switch for all sessions, you can change your default model to gpt-5.1-codex by updating your config.toml configuration file:
model = "gpt-5.1-codex”Today we are introducing a new gpt-5-codex-mini model option to Codex CLI and the IDE Extension. The model is a smaller, more cost-effective, but less capable version of gpt-5-codex that provides approximately 4x more usage as part of your ChatGPT subscription.
Starting today, the CLI and IDE Extension will automatically suggest switching to gpt-5-codex-mini when you reach 90% of your 5-hour usage limit, to help you work longer without interruptions.
You can try the model for a new Codex CLI session using:
codex --model gpt-5-codex-miniYou can also use the /model slash command in the CLI. In the Codex IDE Extension you can select GPT-5-Codex-Mini from the dropdown menu.
Alternatively, you can change your default model to gpt-5-codex-mini by updating your config.toml configuration file:
model = "gpt-5-codex-mini”We’ve shipped a minor update to GPT-5-Codex:
apply_patch.git reset.Codex users on ChatGPT Plus and Pro can now use on-demand credits for more Codex usage beyond what’s included in your plan. Learn more.
You can now tag @codex on a teammate’s pull request to ask clarifying questions, request a follow-up, or ask Codex to make changes. GitHub Issues now also support @codex mentions, so you can kick off tasks from any issue, without leaving your workflow.
Codex is now generally available with 3 new features — @Codex in Slack, Codex SDK, and new admin tools.
You can now questions and assign tasks to Codex directly from Slack. See the Slack guide to get started.
Integrate the same agent that powers the Codex CLI inside your own tools and workflows with the Codex SDK in Typescript. With the new Codex GitHub Action, you can easily add Codex to CI/CD workflows. See the Codex SDK guide to get started.
import { Codex } from "@openai/codex-sdk";
const agent = new Codex();
const thread = await agent.startThread();
const result = await thread.run("Explore this repo");
console.log(result);
const result2 = await thread.run("Propose changes");
console.log(result2);
ChatGPT workspace admins can now edit or delete Codex Cloud environments. With managed config files, they can set safe defaults for CLI and IDE usage and monitor how Codex uses commands locally. New analytics dashboards help you track Codex usage and code review feedback. Learn more in the enterprise admin guide.
The Slack integration and Codex SDK are available to developers on ChatGPT Plus, Pro, Business, Edu, and Enterprise plans starting today, while the new admin features will be available to Business, Edu, and Enterprise. Beginning October 20, Codex Cloud tasks will count toward your Codex usage. Review the Codex pricing guide for plan-specific details.
GPT-5-Codex is now available in the Responses API, and you can also use it with your API Key in the Codex CLI. We plan on regularly updating this model snapshot. It is available at the same price as GPT-5. You can learn more about pricing and rate limits for this model on our model page.
GPT-5-Codex is a version of GPT-5 further optimized for agentic coding in Codex. It’s available in the IDE extension and CLI when you sign in with your ChatGPT account. It also powers the cloud agent and Code Review in GitHub.
To learn more about GPT-5-Codex and how it performs compared to GPT-5 on software engineering tasks, see our announcement blog post.
When working in the cloud on front-end engineering tasks, GPT-5-Codex can now display screenshots of the UI in Codex web for you to review. With image output, you can iterate on the design without needing to check out the branch locally.
codex resume.Learn more in the latest release notes
Codex now runs in your IDE with an interactive UI for fast local iteration. Easily switch between modes and reasoning efforts.
One-click authentication that removes API keys and uses ChatGPT Enterprise credits.
Hand off tasks to Codex web from the IDE with the ability to apply changes locally so you can delegate jobs without leaving your editor.
Codex goes beyond static analysis. It checks a PR against its intent, reasons across the codebase and dependencies, and can run code to validate the behavior of changes.
You can now attach images to your prompts in Codex web. This is great for asking Codex to implement frontend changes or follow up on whiteboarding sessions.
Codex now caches containers to start new tasks and followups 90% faster, dropping the median start time from 48 seconds to 5 seconds. You can optionally configure a maintenance script to update the environment from its cached state to prepare for new tasks. See the docs for more.
Now, environments without manual setup scripts automatically run the standard installation commands for common package managers like yarn, pnpm, npm, go mod, gradle, pip, poetry, uv, and cargo. This reduces test failures for new environments by 40%.
Codex can now generate multiple responses simultaneously for a single task, helping you quickly explore possible solutions to pick the best approach.
Added some keyboard shortcuts and a page to explore them. Open it by pressing ⌘-/ on macOS and Ctrl+/ on other platforms.
Added a “branch” query parameter in addition to the existing “environment”, “prompt” and “tab=archived” parameters.
Added a loading indicator when downloading a repo during container setup.
Added support for cancelling tasks.
Fixed issues causing tasks to fail during setup.
Fixed issues running followups in environments where the setup script changes files that are gitignored.
Improved how the agent understands and reacts to network access restrictions.
Increased the update rate of text describing what Codex is doing.
Increased the limit for setup script duration to 20 minutes for Pro and Business users.
Polished code diffs: You can now option-click a code diff header to expand/collapse all of them.
Now you can give Codex access to the internet during task execution to install dependencies, upgrade packages, run tests that need external resources, and more.
Internet access is off by default. Plus, Pro, and Business users can enable it for specific environments, with granular control of which domains and HTTP methods Codex can access. Internet access for Enterprise users is coming soon.
Learn more about usage and risks in the docs.
Now you can update existing pull requests when following up on a task.
Now you can dictate tasks to Codex.
Added a link to this changelog from the profile menu.
Added support for binary files: When applying patches, all file operations are supported. When using PRs, only deleting or renaming binary files is supported for now.
Fixed an issue on iOS where follow up tasks where shown duplicated in the task list.
Fixed an issue on iOS where pull request statuses were out of date.
Fixed an issue with follow ups where the environments were incorrectly started with the state from the first turn, rather than the most recent state.
Fixed internationalization of task events and logs.
Improved error messages for setup scripts.
Increased the limit on task diffs from 1 MB to 5 MB.
Increased the limit for setup script duration from 5 to 10 minutes.
Polished GitHub connection flow.
Re-enabled Live Activities on iOS after resolving an issue with missed notifications.
Removed the mandatory two-factor authentication requirement for users using SSO or social logins.
It’s now easier and faster to set up code execution.
Added a button to retry failed tasks
Added indicators to show that the agent runs without network access after setup
Added options to copy git patches after pushing a PR
Added support for unicode branch names
Fixed a bug where secrets were not piped to the setup script
Fixed creating branches when there’s a branch name conflict.
Fixed rendering diffs with multi-character emojis.
Improved error messages when starting tasks, running setup scripts, pushing PRs, or disconnected from GitHub to be more specific and indicate how to resolve the error.
Improved onboarding for teams.
Polished how new tasks look while loading.
Polished the followup composer.
Reduced GitHub disconnects by 90%.
Reduced PR creation latency by 35%.
Reduced tool call latency by 50%.
Reduced task completion latency by 20%.
Started setting page titles to task names so Codex tabs are easier to tell apart.
Tweaked the system prompt so that agent knows it’s working without network, and can suggest that the user set up dependencies.
Updated the docs.
Start tasks, view diffs, and push PRs—while you’re away from your desk.
