OpenAI Launches $500 Plan While Cutting the $200 Plan’s Quota on the Same Day: The Trust Cost of Tiered Compute Pricing

On September 29, 2026, OpenAI simultaneously launched the $500-per-month Pro 500 subscription at DevDay and cut the usage quota for new subscribers to its existing $200 plan by half. According to Engadget, the included usage for Work and Codex in Pro 200 fell from 20x Plus quota to 10x, and the weekly message cap for GPT-6 Pro fell from 200 to 100; existing subscribers will temporarily keep their old quotas and later receive one-time compensatory credits.

Rearranging the Three Tiers

After this adjustment, OpenAI's individual subscriptions form a three-tier structure: Pro 100 ($100 per month, 5x Plus quota), Pro 200 ($200 per month, 10x Plus quota for new subscribers), and Pro 500 ($500 per month, 25x Plus quota). According to OpenAI's official wording in its DevDay recap, Pro 500 "offers our highest usage limits, at 25x ChatGPT Plus."

Pro 500's core differentiation comes from Ultrafast—OpenAI's exclusive high-speed inference layer. According to The Decoder, Ultrafast in Codex can reach token generation speeds of up to 300 tokens/second, roughly 8x standard mode; on the API side, it is 6x standard speed. Among individual subscriptions, this speed tier is available only to Pro 500; neither Pro 100 nor Pro 200 can unlock it by purchasing credits, while Enterprise plan users can also use it.

The Dot personal agent debuted the same day, handling delegated tasks such as making calls and sending messages on users' behalf. According to Engadget, all Pro subscribers receive one Dot, and conversations with Dot do not count toward ChatGPT usage limits.

The Quiet Disappearance of Volume Discounts

Behind this adjustment is a less conspicuous but far-reaching change in pricing logic. According to izzedo.chat's unit-price calculation based on OpenAI's official pages: Plus is $20 per month, equivalent to the baseline unit price for 1x Plus quota; Pro 100 is $100 per month for 5x quota, with the unit price still at $20; the old Pro 200 was $200 per month for 20x quota, lowering the unit price to $10; the new Pro 200 is $200 per month for 10x quota, raising the unit price back to $20; Pro 500 is $500 per month for 25x quota, also at a unit price of $20.

In other words, the old Pro 200 was the only tier in OpenAI's subscription system that let heavy users obtain compute at "half price." This quota cut directly eliminates that discount, returning Pro 200 to the same linear pricing level as all other plans. For users who chose Pro 200 for "usage cost-effectiveness," this adjustment changes the core premise of their original purchase decision.

On the API Side: Speed Is the Premium

On the developer side, Ultrafast's pricing logic is more direct. According to The Decoder, API pricing for Ultrafast is $60 per million input tokens and $300 per million output tokens, 6x the standard price. By comparison, GPT-6 Astra's Fast mode is about 2x the standard price and provides about 2.5x speed. This means that when developers face three API speed choices, the costs are roughly: Standard→Fast about 2x, Fast→Ultrafast an additional 3x, for a total of 6x.

This pricing structure has a clear commercial logic for latency-sensitive workloads (such as real-time code generation and long-chain tool calls), but for most asynchronous or batch tasks, the advantage of 300 tokens/second is hard to convert into quantifiable business returns, and the 6x premium lacks sufficient persuasiveness.

Gains and Losses for All Parties

For existing Pro 200 subscribers, they will keep their old quotas in the short term, with one-time credits as compensation later. Users are effectively given two paths: upgrade to Pro 500 for an extra $300 per month, or continue using Pro 200 with half the quota.

For enterprise users, both Pro 500 and the Enterprise tier can access Ultrafast, adding another dimension to procurement decisions for speed-sensitive workflows: they need to assess whether Ultrafast's actual speed gains can cover the 6x API cost, or whether a $500/month subscription can be amortized over sufficient usage.

For competitors, Anthropic, Google, and others have not yet formed a price ladder on individual high-end subscriptions that fully corresponds to OpenAI's three-tier structure. The "compute exclusivity" layer OpenAI built through Ultrafast currently has no direct price benchmark in the market.

Outlook: Signals and Inferences

This move by OpenAI reveals an increasingly clear product strategy: using "speed" rather than "usage" as the core anchor for high-end tiering. The old Pro 200's volume discount was a usage-driven incentive, while Ultrafast is an experience-differentiation premium. The former encouraged users to use more and stay on the platform; the latter encourages users to pay separately for "speed," decoupled from usage. If this shift holds, it means OpenAI may further split pricing dimensions by speed, latency, and exclusive model versions in the future, rather than simply using usage as the sole unit of measurement.

Observable validating signals are: whether new Pro 200 subscriptions decline significantly in the next quarter, and whether early adopters of Pro 500 are mainly enterprise customers. If Pro 200 retention remains stable, it suggests most mid-tier users have strong tolerance for the halved quota and that price elasticity is lower than expected; if Pro 200 churn accelerates, it indicates that the marginal push of this adjustment on mid-tier users has exceeded the stickiness threshold.

OpenAI has not yet disclosed specific impact data of this adjustment on user retention or subscription revenue. The temporary protection window for existing Pro 200 users will expire on October 29 this year, at which point these users' actual reactions will become the first real data points for assessing the impact of this adjustment.