OpenAI Unveils GPT-5-Codex: A Game-Changer in AI-Powered Software Engineering

OpenAI upgrades Codex with a new version of GPT-5 | TechCrunch

OpenAI announced Monday that it's releasing a new version of GPT-5 to its AI coding agent, Codex. The company says its new model, called GPT-5-Codex, spends its "thinking" time more dynamically than previous models, and could spend anywhere from a few seconds to seven hours on a coding task. As a result, it performs better on agentic coding benchmarks. The new model is now rolling out in Codex products -- which can be accessed via a terminal, IDE, GitHub, or ChatGPT -- to all ChatGPT Plus, Pro, Business, Edu, and Enterprise users. OpenAI says it plans to make the model available to API customers in the future. The update is part of OpenAI's effort to make Codex more competitive with other AI coding products, such as Claude Code, Anysphere's Cursor, or Microsoft's GitHub Copilot. The market for AI coding tools has become much more crowded in the last year, as a result of intense user demand. Cursor surpassed $500 million in ARR earlier in 2025 and Windsurf, a similar code editor, was the subject of a chaotic acquisition attempt that saw its team split between Google and Cognition. OpenAI says that GPT-5-Codex outperforms GPT-5 on SWE-bench Verified, a benchmark measuring agentic coding abilities, as well as a benchmark measuring performance on code refactoring tasks from large, established repositories. The company also says it trained GPT-5-Codex for conducting code reviews, and asked experience software engineers to evaluate the model's review comments. The engineers reportedly found GPT-5-Codex to submit fewer incorrect comments, while adding more "high-impact comments." In a briefing, OpenAI's Codex product lead Alexander Embiricos said that much of the increased performance was thanks to GPT-5-Codex's dynamic "thinking abilities." Users may be familiar with GPT-5's router in ChatGPT, which directs queries to different models based on the complexity of a task. Embiricos said GPT-5-Codex works similarly, but has no router under the hood, and can adjust for how long to work on a task in real-time. Embiricos says this is an advantage compared to a router, which decides how much computational power and time to use on a problem at the outset. Instead, GPT-5-Codex can decide five minutes into a problem that it needs to spend another hour. Embiricos said he's seen the model take upwards of seven hours in some cases.

OpenAI unveils new model GPT-5-Codex optimized for 'agentic coding'

OpenAI is rolling out GPT-5-Codex, a new, fine-tuned version of its GPT-5 model designed specifically for software engineering tasks in its AI-powered coding assistant, Codex. The release is part of a broader set of product updates that OpenAI says will help unify its Codex AI assistant across developers' terminals, integrated development environments (IDEs), GitHub, web, and the ChatGPT app. As of today, September 15, GPT-5-Codex becomes the default engine for cloud tasks and code review, with developers able to use it locally through the Codex command line interface (CLI) and IDE extension. The company said it would add the model as an option to the dozens already available in its application programming interface (API) at an unspecified future time, which would enable developers to build their own applications atop it or pipe it into existing ones more easily. Codex has steadily evolved since the company first introduced its CLI in April and web version in May. Two weeks ago, OpenAI consolidated these into a single product experience linked to ChatGPT accounts, enabling developers to switch seamlessly between local and cloud environments without losing context. GPT-5-Codex now extends this unified setup with deeper engineering capabilities. What GPT-5-Codex Brings to the Table Unlike GPT-5, which is built as a general-purpose AI model, GPT-5-Codex is optimized for what OpenAI calls "agentic coding," essentially where the AI agent functions as an autonomous colleague to a developer and can go off and code, run code, and make alterations all on its own for up to hours at a time. It sounds vaguely similar, though distinct enough, from the "agentic swarm" trend that VentureBeat founder Matt Marshall wrote about last week, in which human users can harness multiple agents with defined roles (3 or more) to collaborate on software writing and testing projects all on their own. As for GPT-5 Codex, OpenAI says it was trained on real-world engineering work, from debugging and adding tests to building projects from scratch and executing large-scale refactors. The model is also trained specifically for code review, where it can navigate codebases, reason over dependencies, and run tests to validate correctness. One of the model's key design features is adaptive thinking: that is, GPT-5-Codex adjusts how long it spends reasoning based on the complexity of the task. For short interactions, such as bug fixes or quick Q&A, it responds more quickly than earlier versions. For more demanding projects, it is able to run independently for extended periods -- in testing, it successfully executed refactor jobs lasting more than seven hours, iterating until it delivered a working implementation. But don't get it twisted: GPT-5-Codex does not rely on a router to switch between sub-models. Instead, the model itself dynamically decides how much "thinking" to do at training time. This lets it adjust mid-task if needed, for example revising a plan halfway through execution. For developers, that translates into more consistent handling of both short and long tasks without manual tweaking. In daily use, OpenAI says engineers have relied on Codex for reviewing pull requests, debugging, adding new features, and refactoring existing code. The company reports that GPT-5-Codex has already been integrated into internal review processes, catching hundreds of issues each day. Prior to this announcement, Codex was already running on GPT-5, the general-purpose model, for about the past month. With today's launch, GPT-5-Codex becomes available as a specialized variant trained specifically for coding tasks inside Codex. This marks a shift from a general AI model powering Codex to one purpose-built for agentic software engineering. Updates Across Codex Alongside the new model, OpenAI is releasing enhancements to the broader Codex product: * Codex CLI: Now rebuilt around agentic workflows, the CLI allows developers to attach images like wireframes or screenshots directly, share context, and track progress with built-in to-do lists. It includes simplified approval modes, improved formatting for tool calls and diffs, and support for longer sessions. OpenAI has open-sourced the CLI, which means teams can inspect or extend the harness that connects the model to developer environments. The company sees this as key to helping other organizations align product development with research improvements. * Codex IDE extension: Available in VS Code, Cursor, and similar forks, the extension enables developers to preview local changes and edit code with Codex inside their editor. It maintains context when moving tasks between cloud and local environments, allowing smoother collaboration. According to OpenAI, the IDE extension has quickly become the most popular way to use Codex, reflecting developers' preference for working directly alongside their code. * Codex cloud environment: Behind the scenes, OpenAI has made infrastructure improvements that reduce median completion times by 90 percent. The cloud agent can now auto-configure its environment, install dependencies as needed, and handle screenshots for UI work. Developers can delegate frontend tasks, and Codex will spin up a browser, inspect progress, and attach screenshots to GitHub pull requests. A new feature launching with GPT-5-Codex also enables Codex to output its own screenshots directly within web tasks, giving developers visual confirmation of results without needing to run code locally. * Code review automation: Once enabled in a GitHub repository, Codex automatically reviews pull requests as they progress, flagging issues and even implementing suggested fixes within the same review thread. Developers can also request specialized reviews, such as checks for security vulnerabilities. Codex's review process differs from static analyzers in that it takes a full copy of the codebase, reasons across dependencies, and runs tests before posting feedback -- a level of diligence that even thorough human reviewers may not always apply. Developer Experiences Early customers highlight how Codex is reshaping their workflows. Aaron Wang, Senior Software Engineer at Duolingo, noted that it outperformed other tools in backend code review benchmarks, catching compatibility issues and bugs that others missed. Richard Masters, VP Data & AI at Virgin Atlantic, described how his team can now leave a simple comment in a pull request and Codex will generate a clean update almost instantly, reducing back-and-forth cycles. At Cisco Meraki, Tech Lead Tres Wong-Godfrey used Codex to handle a complex refactor for another team's codebase, freeing him to focus on other priorities while Codex delivered tested, high-quality code. Kevin Royer, Senior AI/ML Engineer at Vanta, said Codex helps him maintain flow by taking on tasks in the background and returning clean diffs. Austin Ray, who leads the AI Dev X team at Ramp, reported that Codex caught an active bug other review tools had missed. Internally, OpenAI teams use Codex across domains like security, frontend, and infrastructure, treating it as a teammate for daily development. Usage of Codex has grown tenfold in the past month, reflecting both external adoption and heavier reliance inside the company itself. Safety and Controls OpenAI emphasizes that Codex runs in a sandboxed environment with network access disabled by default, whether in the cloud or locally. Developers can grant permissions for additional capabilities, such as web search or external tool connections, but are encouraged to review Codex's output before deploying. Each task comes with logs, test results, and citations to help ensure accuracy. The company also classifies GPT-5-Codex as "high capability" in sensitive domains such as biology and chemistry, and has implemented safeguards to minimize associated risks. Procurement teams may note that Codex is not positioned as a replacement for human review but as an additional layer of diligence. OpenAI recommends keeping Codex as a reviewer alongside human engineers, particularly for critical production systems. Pricing and Availability Codex is already bundled with OpenAI's paid subscription ChatGPT Plus ($20/month), Pro ($200/month), Business ($25), as well as Edu and Enterprise plans (no public pricing -- interested parties need to contact OpenAI for pricing). Usage scales by tier: Plus, Edu, and Business plans cover a limited number of focused sessions with the new model per week; Pro supports a full week of development across multiple projects; Business offers add-on credits for expanded use; and Enterprise provides a shared credit pool for large teams. For Plus users, limits are structured around a handful of coding sessions per week, while Pro users should rarely encounter limits in normal use. OpenAI has acknowledged developer feedback on wanting better visibility into usage caps and is considering more transparent indicators in future updates. With GPT-5-Codex, OpenAI positions Codex as more than just an autocomplete or debugging tool. By adapting to tasks of different sizes, working autonomously when needed, and integrating tightly into developer workflows, the company aims to provide a consistent coding partner across environments. Importantly, OpenAI does not intend for Codex to compete with tab-completion tools like GitHub Copilot. Instead, the company views Codex as complementary, with many developers using both -- one for keystroke-level acceleration, the other for higher-level task delegation.

OpenAI Introduces GPT-5-Codex, an AI Model Built Just for Coding

OpenAI has announced its newest model, GPT-5-Codex. The new model has been optimized for agentic coding in OpenAI's suite of AI-powered software engineering tools, which is called Codex. This year, AI programs that can write and edit software have emerged as the most lucrative use case for AI, propelling multiple companies to huge revenue increases. These tools are being used both by professional developers to make their work more efficient, and by casual vibe coders, who lack the technical skill to create websites and apps. The Sam Altman-led company claims that by training this new AI model on real-world engineering tasks, it can outperform the default model. In a benchmark that compared that model and GPT-5-Codex's ability to refactor code (essentially reorganizing and cleaning up code), GPT-5-Codex scored nearly 20 percent higher than the default model, which is simply called GPT-5. GPT-5-Codex is also said to be a strong independent worker. It can work autonomously on software for long stretches of time. According to a press release, OpenAI has seen the model "work independently for more than seven hours at a time on large, complex tasks, iterating on its implementation, fixing test failures, and ultimately delivering a successful implementation." The new model could also help alleviate one of the most notable pain points of vibe coding: bad code. Many software developers have remarked that much of their time working with AI-assisted code editors is spent cleaning up the AI's code, which isn't always as thoughtfully written as a human expert's would be. But OpenAI says that GPT-5-Codex has been "trained specifically for conducting code reviews and finding critical flaws." In practice, the company says, this means GPT-5-Codex will review an entire codebase to identify flaws and autonomously test apps to find errors. OpenAI says that Codex currently handles "the vast majority" of proposed changes to code being written by OpenAI staffers, "catching hundreds of issues every day -- often before a human review begins." But even with its improved code review abilities, OpenAI still recommends using Codex as an additional reviewer; it says in a press release that it is "not a replacement for human reviews." Unlike the normal version of GPT-5, GPT-5-Codex won't be immediately available via API, and OpenAI recommends only using the model for coding tasks in Codex-supported environments. In addition, Codex is coming to mobile devices for the first time. Previously, in order to access Codex, you'd either need to use ChatGPT on a desktop computer or invoke Codex in an IDE (integrated development environment) like VSCode or Cursor. Now, Codex will be accessible in the ChatGPT iOS app, enabling easier coding on the go. Codex, and GPT-5-Codex, is available across all of ChatGPT's paid tiers, with $20-per-month ChatGPT Plus members getting enough access to "cover a few focused coding sessions each week." Meanwhile, $200-per-month ChatGPT Pro members will get enough to "support a full workweek across multiple projects." Companies that pay for ChatGPT's SMB-focused Business plan can purchase credits to give their developers more access to Codex, while larger companies with ChatGPT's Enterprise plan get a shared credit pool. In OpenAI's press release, engineers and tech leads at companies including Cisco, Duolingo, Ramp, Vanta, and Virgin Atlantic praised Codex's utility, but it remains to be seen if GPT-5-Codex can help OpenAI take market share away from Anthropic, whose similar Claude Code product has proved very popular with professional and casual software developers. The extended deadline for the 2025 Inc. Best in Business Awards is this Friday, September 19, at 11:59 p.m. PT. Apply now.

OpenAI releases GPT-5-Codex for enhanced software engineering By Investing.com

Investing.com -- OpenAI launched GPT-5-Codex on Monday, a specialized version of GPT-5 optimized for agentic coding in Codex with enhanced capabilities for software engineering tasks. The new model excels at both quick interactive sessions and independently handling complex, lengthy tasks. It also features code review functionality designed to identify critical bugs before software deployment. GPT-5-Codex is now available across all Codex platforms, serving as the default option for cloud tasks and code review. Developers can also select it for local tasks through Codex CLI and the IDE extension. OpenAI has been steadily enhancing Codex since launching Codex CLI in April and Codex web in May. Two weeks ago, the company unified Codex into a single product experience linked to users' ChatGPT accounts, allowing seamless work transitions between local environments and cloud platforms without losing context. The platform now functions across multiple development environments including terminals, IDEs, web browsers, GitHub, and the ChatGPT iOS app. Codex is included with ChatGPT Plus, Pro, Business, Edu, and Enterprise subscription plans. According to OpenAI, these updates advance their vision of creating a coding teammate that understands user context, works collaboratively, and reliably handles team tasks. This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.