OpenAI Announces Three-Tier Overhaul of ChatGPT Pro Subscription

On September 29, 2026, in conjunction with DevDay 2026, OpenAI restructured the ChatGPT Pro subscription into three tiers: Pro 100, Pro 200, and Pro 500. This overhaul reduces usage limits for existing Pro 200 subscribers, while the top-tier Pro 500 is the only plan to include the Ultrafast mode for GPT-6 Astra.

Key Changes by Subscription Tier

• Pro 100: $100/month, 5x usage compared to Plus, Ultrafast not included
• Pro 200: $200/month, usage reduced to 10x compared to Plus (existing subscribers retain 20x until October 29, 2026), Ultrafast not included
• Pro 500: $500/month, 25x usage compared to Plus, includes Ultrafast mode for GPT-6 Astra (exclusive among individual plans)

Common to all tiers are monthly billing only, GPT-6 Astra-based Pro inference, always-on agent dot, unlimited image generation, and maximum-level Deep Research and Memory features. The context window for GPT Reasoning models is 400K tokens.

Price Efficiency Changes and Impact on Existing Users

All plans (Plus, Pro 100/200/500) will now apply the same unit price ratio of one Plus unit per $20, eliminating the previous "bulk purchase discount" effect of Pro 200. Tier selection will now be determined by total usage needs and the requirement for Ultrafast, rather than unit price.

Existing Pro 200 subscribers can maintain their previous usage levels (20x) until October 29, 2026, after which they will transition to the new usage limits (10x). OpenAI stated that it plans to issue significant additional credits to existing subscribers. New sign-ups for Pro 200, which had been suspended since September 10, 2026, will resume.

Detailed Analysis of Ultrafast Mode

Ultrafast is a premium service tier that runs on faster inference infrastructure without changing the model. It offers up to 8x faster token generation speeds in Codex (300 tokens per second) and up to 6x faster speeds in the API. However, this refers to token generation speed and is not equivalent to total task completion time or billing multipliers.

• Supported Models: GPT-6 Astra (GPT-6.1 Sol to be added soon)
• Usage Consumption: Consumes 8x the usage compared to standard mode. If the 25x usage allowance of Pro 500 is used exclusively for Ultrafast, it equates to approximately 3x the workload of Plus.
• API Pricing: Input $60.00 and output $300.00 per million tokens, which is 6x more expensive than Standard.
• Restrictions: Exclusive to Pro 500 among individual plans; for Enterprise, workspace owners must configure permissions. It is unavailable in workspaces requiring data residency outside the United States.