OpenAI launches $500/month Pro tier with Ultrafast mode, halves $200 allowance

OpenAI used its DevDay 2026 event on September 29 to introduce a new Pro plan priced at $500 per month, its most expensive consumer subscription yet. The tier carries the highest usage allowance of any OpenAI subscription and adds access to Ultrafast, a premium speed mode for workloads where latency matters most. The same day, the company also announced GPT-6.1 Sol, a model it says matches much of flagship GPT-6.0 Astra's performance in coding and computer use at a fifth of the price.

What Ultrafast changes

According to OpenAI, Ultrafast lets its Codex and Work applications generate up to 300 tokens per second, and The New Stack reports that the mode promises up to eight times the standard token speed in Codex. OpenAI describes it as a premium speed tier for workloads where speed is critical, and says the same tier is available to API customers. Ultrafast support for the new GPT-6.1 Sol model is promised for a later date.

The $500 plan includes 25x the usage allowance of the $20 ChatGPT Plus subscription, though it is not yet clear what that means in absolute terms. Every Pro subscriber also receives a single Dot, OpenAI's new personal AI agent, and conversations with a Dot do not count against the ChatGPT usage limit.

The $200 tier gets smaller

The same announcement cuts the value of the existing $200 Pro plan. From October 30, the included usage for ChatGPT Work and Codex falls from 20x the Plus allowance to 10x, while the weekly GPT-6 Pro message cap in ChatGPT drops from 200 to 100. Existing subscribers keep their current limits through October 29, after which they receive a one-time grant of 62,500 usage credits, worth $2,500 according to OpenAI, that expire on December 31, 2026. OpenAI says it will not bring back the old five-hour limit, so subscribers can spend their weekly allowance at their own pace.

Why it matters

The 20x multiple for a top-tier plan had long been the de facto standard across frontier labs, including Google's $200 Gemini AI Ultra plan and Anthropic's $200 tier, while most labs also offer a $100 plan at 5x. The change, first previewed by head of product and platform Tibo Sottiaux on X, drew negative reactions from developers. Sottiaux argues subscribers will still get more work done than they did a month ago because the underlying models are cheaper and more efficient. The New Stack notes that is not what users signed up for when they bought the $200 plan.

Sources: Engadget · The New Stack