What OpenAI Actually Released

On May 14, 2026, OpenAI rolled out Codex inside the ChatGPT mobile app — available now in preview on iOS and Android across all plans, including the free tier, according to OpenAI's official product blog.

The release is not a full coding environment on your phone. Your phone acts as a remote control. Codex runs on your Mac, a Mac mini, a company devbox, or a remote server. The mobile app connects to that machine through a secure relay layer and pulls live state — terminal output, screenshots, test results, diffs, and pending approvals — according to OpenAI. Your credentials, files, and local setup stay on the host machine. Nothing sensitive travels to your phone.

Windows support is coming. The current release only connects to macOS.

By the Numbers

OpenAI says more than 4 million people now use Codex every week, according to Thurrott. That's the first hard user figure the company has attached to the product since it launched approximately a year ago.

The Competitive Reality Mainstream Coverage Is Soft-Pedaling

Every outlet covered the launch. Most buried the competitive context.

TechCrunch noted Anthropic released its "Remote Control" feature in February 2026 — months before OpenAI shipped the same capability. The Verge mentioned Claude Code's "surge in popularity" as a driver behind OpenAI's recent sprint.

OpenAI is catching up.

The Verge reported that OpenAI has been cutting "side quests" — redirecting resources toward enterprise growth and Codex development specifically because Claude Code has been winning market share among developers and businesses. A company changing its roadmap under competitive pressure is rarely a minor detail.

What OpenAI Is Actually Building

Mobile access is one piece of a larger expansion.

In March 2026, OpenAI confirmed it is building a single desktop "superapp" that combines ChatGPT, Codex, and its Atlas browser into one experience, according to Engadget. Earlier this month, the company also released a Chrome extension so Codex can operate inside live browser sessions, per TechCrunch. Last month, Codex gained the ability to run in the background on desktop environments autonomously.

Mobile access closes the loop — a developer starts a task at the office, leaves, and Codex keeps working. When it hits a decision point, it pings your phone. You approve or redirect. Work continues.

The architecture is well-designed. The secure relay keeps trusted machines reachable across devices without direct public internet exposure, according to OpenAI. That matters for enterprise customers who aren't going to route their production codebase through some cloud middleman with undefined security practices.

What This Means for Developers

If you're a developer already using Codex, this is straightforwardly useful. Long-running agentic tasks no longer require you to stay tethered to your desk. You can start a bug investigation from your phone while standing in a coffee line, according to OpenAI's own use-case examples.

For free-tier users, this is notable. Full Codex mobile access at zero cost is aggressive pricing. OpenAI is clearly prioritizing user growth over near-term monetization on this feature.

The Competitive Context

Most outlets framed this as OpenAI "expanding" or "growing" its product. Clean, positive, no friction.

The reality is different: Claude Code has become the preferred agentic coding tool among a significant slice of professional developers. OpenAI knows this. The internal restructuring — cutting side projects, pushing enterprise — reflects a company responding to a competitor that moved faster.

Codex is a real product with real user numbers and solid technical execution. But the timing tells the story. Anthropic had this first, and OpenAI is closing the gap as fast as it can. The race to build the dominant AI coding agent is wide open, and neither company has won it yet.