OpenAI’s Codex Hits 4M Weekly Users and Gets a Proper Enterprise Push

OpenAI just dropped some numbers that caught my attention: Codex now has 4 million weekly active users. That’s not just developers tinkering with autocomplete in their personal repos — that’s actual sustained usage. And to back that up, they’re launching Codex Labs, a new program aimed squarely at the enterprise, with partners like Accenture, PwC, and Infosys.

Let’s be real for a second. Codex has always felt like one of those tools that’s amazing in a demo but messy in production. I’ve seen teams try to drop it into their CI/CD pipeline and end up with more broken builds than shipped features. The problem isn’t the model — it’s the integration. Enterprise codebases are tangled, permissions are a nightmare, and nobody wants an AI committing code that bypasses review.

Codex Labs seems designed to fix exactly that. Instead of just selling API credits, OpenAI is partnering with the Big Four consultancies and systems integrators who already have their hands in enterprise infrastructure. These are the firms that know how to navigate compliance, legacy systems, and the dreaded “we’ve always done it this way” mentality. Accenture, PwC, Infosys — they’re not flashy, but they’re effective at getting enterprise software deployed.

The 4M WAU figure is higher than I expected. For context, that’s roughly the size of a mid-tier SaaS product’s user base. It suggests Codex is moving beyond early adopters and into teams where code quality actually matters. I’d love to see the breakdown between individual developers versus team/org accounts, but OpenAI isn’t sharing that yet.

What I’m most curious about is how Codex Labs handles the “last mile” problem. Enterprise adoption isn’t about the demo — it’s about training data, fine-tuning on internal codebases, and ensuring the model doesn’t hallucinate proprietary logic. The partnership model suggests OpenAI knows they can’t do that alone. Good move.

There’s also the question of pricing. Enterprise deals with consultancies usually mean per-seat licensing or project-based fees, not the consumption-based pricing developers are used to. If Codex Labs pushes a model where the cost is baked into a larger digital transformation contract, it could actually accelerate adoption. Or it could make Codex feel like just another expensive add-on that gets cut in the next budget review.

One thing I appreciate: OpenAI didn’t overhype this. No claims about replacing developers or achieving AGI through autocomplete. Just a straightforward announcement about scaling a tool that’s already working. That’s refreshing.

I’ll be watching to see if Codex Labs actually publishes case studies with metrics — like reduction in cycle time, fewer bugs in production, or developer satisfaction scores. That’s the kind of data that sells enterprise software, not press releases.

For now, 4M WAU and partnerships with the usual suspects is a solid bet. Let’s see if they execute.