OpenAI is building a ChatGPT tier that costs $500 a month, more than double its current $200 Pro plan. TestingCatalog, an app-teardown outlet, found the pricing reference inside unreleased subscription code and reported it on September 24, 2026. The listing describes the new plan, called Pro Max, mainly through access to what OpenAI’s own wording calls “Fastest Work and Codex.”
That phrasing matters more than the price tag. Work is OpenAI’s tool for long-running tasks, and Codex handles coding jobs that can run for hours or days. Both are the kind of workload where a faster model saves real time, unlike a quick chat answer where speed barely registers. The leak does not confirm whether Pro Max also raises usage caps beyond that speed boost.
TestingCatalog’s reporting ties the “Fastest” label to Cerebras, the chip maker OpenAI already uses for low-latency inference. OpenAI ran a preview of an “Ultrafast” mode on its GPT-5.6 Sol model in August, claiming up to 14 times faster output for a limited set of customers, though it never named Cerebras directly. TestingCatalog has not confirmed that Pro Max draws on the same infrastructure, and OpenAI has not addressed the plan publicly.
The timing lines up with OpenAI’s developer conference on September 29, where the company is expected to cover APIs, developer tooling and new subscription tiers. OpenAI has also paused new sign-ups for its existing $200 Pro plan because of system load, a detail that points toward a capacity problem rather than a pure upsell. Selling a faster lane at a much higher price is one way to relieve pressure on the standard tier without expanding it.
A $500 plan would price ChatGPT well above every comparable subscription on the market. Anthropic and Cursor both cap their top individual tiers near $200, and Google’s AI Ultra plan runs about $250. If OpenAI holds this price, it is not competing with those plans on value. It is betting that developers running Codex on multi-day jobs will treat the fee as a compute cost rather than a software subscription, closer to renting a faster GPU than paying for a chat app.
The open question is what OpenAI is actually charging for. A subscription justified by raw speed is a different business than one justified by higher usage limits, and the two invite different customers. Speed pricing suggests OpenAI is passing through the cost of premium inference hardware to the users who need it most. Usage pricing suggests it still sees room to grow revenue per seat without changing the underlying model access. Nothing in the leaked code settles that, and OpenAI has not commented.
Teams already paying $200 a month for Codex-heavy workflows should watch DevDay on September 29 for the actual terms before assuming Pro Max is a straightforward upgrade.
Reported by TestingCatalog on 24 September 2026.
