OpenAI Codex in 2026: Status, Timeline and 8 Million Users

Short answer first, because it’s the thing people actually search for: Codex was not discontinued. It is the fastest-growing product OpenAI has shipped since ChatGPT itself, and in 2026 it went from 600,000 weekly users to 8 million.

The longer answer is more interesting, because that 8 million number hides something worth reading carefully.

Codex in 2026: the timeline

• February 2026 — Codex ships as a desktop app running GPT-5.3-Codex. Not an IDE plugin, not a flagged experiment: dock icon, automatic updates, a real product.
• March 5, 2026 — GPT-5.4 lands. The release cadence for the Codex line becomes quarterly, sometimes monthly.
• April 21, 2026 — 4 million weekly active users.
• June 2, 2026 — Codex passes 5 million weekly active users on its own, as reported by Constellation Research.
• July 9, 2026 — GPT-5.6 launches and the curve goes vertical.
• July 12–14, 2026 — 6 million, then 7 million roughly 24 hours later, then 8 million, according to The New Stack.

Start of year to mid-July: 600,000 to 8 million. That’s not a product finding traction. That’s a category forming.

The number everyone quotes wrong

Here’s the part worth being precise about, and almost nobody is.

The clean, standalone metric — Codex by itself, weekly active users — is the 5 million figure from early June. The 8 million headline from July counts Codex and ChatGPT Work together. Bigger number, looser definition.

I spent two decades building enterprise search before I pivoted to agents, and this is a pattern I’ve seen at every hype peak: the metric quietly widens as the story gets told. It doesn’t mean anyone is lying — combined usage is a legitimate thing to measure. It means that if you’re putting this in a board deck to justify budget, cite 5 million standalone and you’ll never have to walk it back.

Who is actually using it

The most underrated fact in the whole dataset: knowledge workers are now roughly 20% of users, and they’re growing more than three times faster than developers.

Read that again. The fastest-growing segment of a coding agent is people who don’t write code for a living. That’s the same transition search went through — a tool built for specialists becomes infrastructure for everyone, and the specialists are the last to notice.

Speed mattered more than the model

One of the less-discussed announcements was technical, not product: Spark, running on Cerebras infrastructure, got roughly 15x faster. That’s not an incremental latency bump — it’s the kind of jump that changes how you use the tool day to day. When the “ask, wait, review” cycle drops from seconds to fractions of a second, the developer stops context-switching and starts working inside the agent’s loop.

It’s the same lesson I learned building search engines: low throughput kills adoption before result quality ever gets a chance to matter.

The stamp that unlocks corporate budget

OpenAI positioned itself as a leader in Gartner’s Magic Quadrant for Enterprise AI Coding Agents (openai.com). Enterprise analysts are conservative by nature — they don’t hand out leader positions without evidence of adoption inside corporate environments, not just individual devs poking at a side project.

That stamp is exactly the signal CTOs use to justify budget internally. When Gartner validates a category, the corporate buying cycle unlocks.

What this means if you’re hiring agents, not just tools

Put it together — a mature desktop product, infrastructure 15x faster, 5 million standalone weekly users, non-developers as the fastest-growing segment, and enterprise-analyst validation — and the message is simple: coding agents stopped being a bet. They became infrastructure any sensible company is already evaluating.

The question left isn’t “should I use AI to code.” It’s “what agent architecture do I build on top of this, and how do I keep it from becoming a fragile dependency on a single vendor.” Model versions shipped monthly in 2026. Anything you build has to treat the model as a living dependency, not a foundation.

That’s exactly the decision we work through with companies that come to Arvor for AI agent consulting or to structure a corporate brain with BRAIN MAKER. If your company still treats Codex as an individual developer’s experiment, you’re a cycle behind — and in 2026 a cycle lasted about six weeks.