On 29 September 2026, before 2,500 developers at DevDay, OpenAI reshaped the top of its price pyramid. Pro 500 arrives at $500 a month, the only Pro plan with Ultrafast: up to 300 tokens per second in Codex, up to eight times standard speed, per The Next Web.
At the same time, Pro 200—reopened to new subscribers—will see included usage cut. From 30 October, the ChatGPT Work and Codex allowance falls from 20× to 10× the Plus allotment; GPT-6 Pro chat messages fall from 200 to 100 a week. OpenAI pledged not to reintroduce Pro 200’s five-hour limit that once frustrated users.
Current subscribers keep their limits through 29 October and receive a one-time usage credit worth $2,500 that expires 31 December 2026. It is a temporary buffer for an effective price rise per unit of capacity.
The DevDay package also adds ChatGPT Space (replacing Library) with “pages” editable by teammates, ChatGPT and dots; collaborative slides within weeks; @ChatGPT in Slack and Teams without every teammate needing a licence; and a Mac Meetings plugin in beta that takes notes without a visible bot—verbal consent required, though the in-app reminder is visible only to the user.
Codex gains cloud tasks that keep running while a laptop sleeps, a voice-steerable CLI, desktop Code Review and Codex Security Cloud with Daybreak Blue models. Sign in with ChatGPT lets Plus/Pro users spend plan usage in 16 partner tools (Devin, Notion, Vercel) with weekly caps per app.
The economic read is clear: agency (dots, computer use, Ultrafast) concentrates on the $500 tier, while $200 becomes a less generous middle plan. With 1.2 billion weekly ChatGPT users and over 35 million using Work/Codex weekly, OpenAI is prioritising margin and GPU capacity control.
For freelancers and startups living on Pro 200, the cut forces an audit: what share of the quota went to Codex batch jobs? Does Pro 500 pay off, or a partial shift to API with hard budgets? The $2,500 credit buys time, not a permanent policy.
The U.S. enterprise Marketplace lets eligible customers put part of OpenAI spend commitments toward approved software (Adobe, Figma, Salesforce, CrowdStrike among 32 betas). Private Intelligence promises automated safety reviews without OpenAI staff seeing content; Private Inference is due this autumn.
Politically, raising Pro prices while signing safety pacts and cancelling models sends a dual message: more economic friction may reduce abuse at scale, but it also reserves agentic autonomy for those who pay like a company. Competition regulators will watch whether packaging harms rivals that depend on open APIs.
EEA, Swiss and UK users, moreover, do not get dots for now: they pay global prices without the flagship agentic feature. That geographic asymmetry will feed consumer complaints and AI Act questions on availability and transparency.
Neural Pulse advises teams to measure real tokens/week before 30 October, set quota alerts and decide whether Ultrafast justifies an extra $300 or whether API calls with caching and batches are more predictable. DevDay’s “new renaissance” marketing copy does not pay the GPU bill.
The close: Pro 500 is not just an SKU; it is a statement that speed and always-on agents are a rationed luxury. In the same week SoftBank issues at 9.75% to fund OpenAI, users feel the echo on their credit cards.
