OpenAI has quietly been losing ground in one of the most strategically important corners of AI: professional developer tooling. On Thursday, the company moved to claw it back, announcing a new $100-per-month ChatGPT Pro tier built specifically around Codex, its agentic coding tool. The move plugs a glaring gap in OpenAI's pricing stack and puts Anthropic squarely in the crosshairs.
Until now, the jump from ChatGPT Plus ($20/month) to the full Pro plan ($200/month) was steep enough to lose serious developers to rivals who offered a more sensible middle rung. Anthropic has had a $100 "Max" tier for Claude for some time. OpenAI just matched it — and is loudly arguing it offers more coding capacity per dollar.
What OpenAI Is Actually Offering
The new $100 Pro tier delivers five times the Codex usage of the $20 Plus plan, targeting developers who run heavy, multi-hour coding sessions rather than casual daily use. The existing $200 plan isn't going anywhere — OpenAI confirmed it's still available even though it's been removed from the public pricing page — but the $100 tier now serves as a more accessible on-ramp for power users who don't need the $200 plan's 20x limits.
To make the math concrete: the new $100 tier gives you roughly 450–2,250 local messages and 100–600 cloud tasks every five hours when using GPT-5.3-Codex, plus 100–250 pull request code reviews per week. Those aren't soft suggestions — actual limits vary based on task complexity, codebase size, and context length. OpenAI is also reportedly rebalancing how Plus users consume Codex, distributing capacity more evenly across the week rather than allowing it to be burned through in a single burst. Read that carefully: for existing Plus subscribers, daily heavy usage may feel tighter going forward.
There's also a catch worth flagging for early adopters. OpenAI is offering elevated Codex limits on the new $100 tier through May 31. Developers who sign up, run their projects hard, and never hit a ceiling should expect that experience to change after that date.
The Real Fight: Anthropic's Momentum
The $100 tier isn't just about pricing hygiene. It's a competitive response to a rival that has been growing faster than OpenAI in the segment that matters most for long-term monetization: professional developers and enterprises.
Anthropic recently disclosed an annualized revenue run rate north of $30 billion — reportedly ahead of OpenAI's $24–25 billion — and much of that momentum traces back to Claude Code's grip on the developer community. Claude Code has become a default tool for many engineers doing heavy agentic work, and its $100 Max plan gave those users a natural home.
OpenAI's counter-argument, delivered somewhat pointedly through a spokesperson, is that Codex delivers more coding capacity per dollar across paid tiers, "with the difference showing up most clearly during active coding use." Whether that holds up under real-world workloads is something developers will test immediately.
There's a broader subplot here that makes the $100 launch feel more deliberate. Anthropic recently blocked Claude subscriptions from powering third-party agentic tools like OpenClaw, forcing those users onto the API with separate billing. That decision fractured a community of power users who had been using OpenClaw as their coding harness of choice. OpenAI hired OpenClaw's creator, Peter Steinberger, in February and has since made a point of emphasizing that its tools don't carry those same restrictions. The $100 tier, timed as it is, reads as an explicit invitation to displaced OpenClaw users looking for a new home.
This puts pressure on Anthropic to either restore flexibility for third-party integrations or accelerate its own first-party tooling fast enough to retain developers who are now weighing their options.
Usage Numbers Suggest Real Traction
One data point OpenAI shared is worth sitting with: more than 3 million users per week are now using Codex globally, a figure that has grown roughly 5x over the past three months with month-over-month growth above 70%. That's a strong signal that the underlying product has genuine pull — and that the limiting factor for converting casual users into paying subscribers may have been price architecture rather than product quality.
Sam Altman framed the new tier straightforwardly on X: it's launching "by very popular demand." That's not just marketing. A $100 price point for heavy Codex use was an obvious gap, and the speed of the existing growth suggests OpenAI may have been leaving significant revenue on the table by not offering it sooner.
What This Means
- For developers: If you're doing serious daily Codex work and have been sitting on Plus or paying $200 for features you don't use, the $100 tier is worth evaluating immediately — just be aware the generous early limits expire June 1. Run your own benchmarks against Claude Code before committing.
- For founders and teams: OpenAI's move widens the competitive gap between it and tools like GitHub Copilot and Google's Gemini-powered offerings, which haven't matched this kind of tiered agentic capacity at the $100 price point. Teams budgeting for AI coding tools now have a clearer decision matrix.
- For Anthropic: The pressure isn't existential, but it's real. Anthropic's restriction of third-party Claude integrations was a calculated economic decision, but it created an opening OpenAI has now explicitly targeted. If Claude Code's quality advantage narrows as Codex improves, that tradeoff looks worse in hindsight.
- For the AI market broadly: Subscription pricing for agentic developer tools is consolidating around a $100-per-month benchmark. The product war is increasingly being fought not on raw model capability but on usage limits, task throughput, and how gracefully tools handle long, complex sessions. That's a different race — and one where infrastructure efficiency matters as much as model quality.
OpenAI didn't just add a pricing tier. It restructured the competitive landscape for professional AI coding tools in a single announcement. The developers who stand to benefit most are the ones willing to stress-test the limits before May 31.
