OpenAI’s super app is here: Codex and ChatGPT merged, launching within weeks

```

OpenAI is merging its most successful programming tool, Codex, with ChatGPT into a desktop "super app," scheduled for release in the coming weeks. This integration marks a major strategic shift for OpenAI—Codex has evolved from a developer-oriented programming tool into a universal productivity platform for all knowledge workers.

On June 2, according to tech media The Information, at a corporate event held Tuesday, OpenAI announced its super app integrating ChatGPT, Codex, and its browser tool Atlas, which will be launched soon. OpenAI’s leadership stated that the core reason for the merger is that Codex performs better than ChatGPT on many tasks—especially in handling long, multi-step tasks and invoking external tools. Thibault Sottiaux, head of the Codex team, noted Codex is better than ChatGPT at tasks like editing complex spreadsheets that require coding.

On the user growth front, Codex is surging. As of the end of May, Codex’s weekly active users have exceeded 5 million, up from 4 million the previous month and 3 million two weeks earlier—a staggering pace. OpenAI co-founder and president Greg Brockman revealed last month that Codex’s enterprise client revenue increased by 50% week over week; CEO Sam Altman also said at a staff meeting that Codex’s overall usage grows about 5% daily. This momentum is driving OpenAI to accelerate Codex’s capabilities for its over 900 million consumer users, expanding revenue before its IPO.

The timing of the super app launch is delicate. Anthropic secretly submitted its IPO application Monday, with enterprise client revenue accounting for over 80% and annualized revenue exceeding $47 billion. By contrast, OpenAI’s enterprise client revenue percentage is still above 40%, and Chief Revenue Officer Denise Dresser’s goal is to raise that to 50% by year-end.

ChatGPT and Codex Merge: The Strategy Behind the Super App

The merger of ChatGPT and Codex is one of OpenAI's most important recent product decisions.

Reportedly, the direct reason for the integration is Codex’s key advantages in technical architecture that ChatGPT lacks. Sottiaux pointed out that Codex has a superior "harness"—the underlying software framework that enables AI agents to call tools and execute operations on behalf of users—which is a core reason for the merger.

Currently, there is a fundamental technical difference between ChatGPT and Codex: ChatGPT runs in the cloud, while Codex runs locally on user devices, directly accessing local files. Sottiaux admitted that how to combine these two approaches is one of his biggest current challenges.

OpenAI’s upcoming plan is to integrate Codex’s capabilities into ChatGPT and allow users to choose whether Codex or ChatGPT responds to their queries.

OpenAI’s Head of Enterprise Products, Alexander Embiricos, said this design is meant to help users gradually understand the more complex tasks Codex can handle. In the long run, users will not need to select tools manually—the model will automatically decide whether tasks should be executed locally or in the cloud.

This merger is also part of OpenAI's broader organizational restructuring. Last month, OpenAI unified the ChatGPT, Codex, and API teams into a single organization, led by Sottiaux and responsible for "core products and platforms."

At Tuesday’s demo, OpenAI’s product leaders showcased the new product design—"plugins" integrating with Canva and Figma. The timing was meaningful: Anthropic recently faced criticism from partners after launching its own design features, as it was seen as competing with ecosystem design tools.

OpenAI’s move intends to send a very different signal: The platform will not copy partners’ core businesses, but instead adopt open integration to empower them.

At the event, Sam Altman directly stated this position: "We have no intention of competing with your businesses." This marks a clear shift from previous negative perceptions of OpenAI competing head-to-head with vertical AI startups.

Beyond design, OpenAI also released special features for professional scenarios like investment banking, and introduced new capabilities allowing enterprise clients to quickly create interactive websites (such as dashboards and planning tools), replacing traditional documents or slides to improve information-sharing efficiency.

Enterprise users can also fine-tune Codex's output results in spreadsheets or presentations.

Codex Rising: From Programming Tool to Universal Productivity Platform

Codex’s rise in strategic importance comes from OpenAI's deep reflection on the competitive landscape.

According to The Information, as early as autumn 2024, Anthropic’s Claude had already surpassed OpenAI’s models on some internal metrics, causing OpenAI to take notice—since OpenAI has long believed AI programming ability is a key path to accelerated research and achieving superintelligence. When Anthropic launched the Claude Code research preview in February 2025, OpenAI realized the gap could not be ignored, prompting it to form a dedicated team for AI programming tools.

This team operated as a "startup within a startup." Sottiaux noted, the team open-sourced Codex’s core code to more efficiently gather user feedback and suggestions—something rare in most companies. Embiricos added that Codex's edge in some areas is partly because the same team simultaneously developed both the model and the product, whereas many other OpenAI products have separate teams for application and research.

Major releases drove Codex’s stepwise user growth. Sottiaux said the GPT-5.2 release in December 2024 greatly improved Codex’s ability to handle long tasks; GPT-5.5, launched in April 2025, was OpenAI’s first flagship AI since GPT-4.5 in February last year to introduce a new base model, resulting in a qualitative leap.

DJ Sampath, Cisco’s SVP of AI Software & Platforms, said GPT-5.5 greatly reduced dependence on manual intervention for long tasks. Basis, a New York AI accounting startup, co-founder Mitch Troyanovsky put it: "Codex improved since GPT-5, but with GPT-5.5 the gap became impossible to ignore."

Developer Battle: Codex Surpasses Claude Code

Codex’s rapid growth largely stems from winning back the developer community.

Several developers said they switched to Codex after OpenAI launched its desktop app in February. Previously, developers primarily used Claude Code via terminal command line, but the desktop app format is better for non-tech users and helps developers manage multiple coding agents simultaneously. Notion’s product lead Max Schoening noted he likes to start coding tasks on his computer, then monitor progress through Codex’s newly released mobile app.

Developers’ opinions on the tools are clearly divided. Many said Codex is stronger when engineers have clear instructions, while Claude Code is better when ideas are unclear and the model needs to explore solutions itself—because Claude Code excels at inferring intent when user descriptions lack precision.

Notion software engineer Varun Rau said:

"If I have no idea what the solution looks like and need the model to figure it out, Claude Code feels a bit better; but when I have clear specs, I trust Codex to execute."

He also noted Codex is "obsessive" about bug fixing—he and colleagues once spent weeks before realizing there was a bug in their code, because Codex would auto-fix it every time the server was started.

However, developers also pointed out limitations: Claude Code’s autonomy sometimes causes it to deviate from instructions, while Codex’s obsessiveness may make it track and fix a lot of bugs even when only small changes are needed.

Commercially, competition between OpenAI and Anthropic is reaching fever pitch, with both facing urgent pressure to expand revenue before listing.

Reportedly, most of Codex’s 5 million weekly active users are paid users, but a sizable proportion remains unpaid. By comparison, Anthropic’s Claude Code and Claude Cowork are only available to paying users and do not disclose user numbers.

In terms of revenue, Anthropic currently holds an advantage. Claude Code’s strong performance drove Anthropic’s annualized revenue to surpass $47 billion last month, up fivefold from the start of the year. OpenAI disclosed $25 billion in annualized revenue in March this year, but hasn't updated since; The Information previously reported OpenAI's annualized revenue surpassed $30 billion.

Analysts pointed out that one of OpenAI’s strategic goals with the super app is to bring Codex's capabilities to its 900 million+ consumer users, greatly increasing revenue. Meanwhile, the integration of Codex and ChatGPT is viewed as a key step for OpenAI to build a universal automation agent for knowledge workers—this is very similar to Anthropic’s strategy of extending Claude Code with Claude Cowork. Both companies are rushing toward IPO to secure large-scale capital needed for AI chip purchases and hiring researchers.

Risk Notice and DisclaimerThe market has risks; invest cautiously. This article does not constitute personal investment advice, nor does it take into account any individual user's special investment objectives, financial situation, or needs. Users should consider whether any opinion, viewpoint, or conclusion in this article fits their specific circumstances. Investments are made at your own risk. ```