ChatGPT Pro 500 Launches at $500 With 25x Plus Usage and Astra Ultrafast, While Pro 200 Drops to 10x

OpenAI announced Pro 500, a $500-a-month ChatGPT plan, at DevDay 2026 on September 29. OpenAI says it "offers our highest usage allowance at 25 times the ChatGPT Plus allowance and includes access to Ultrafast," the new speed tier that launched for GPT-6 Astra in ChatGPT Work and Codex (OpenAI DevDay recap). Ultrafast now supports GPT-6 Astra and GPT-6.1 Sol (Codex speed docs). OpenAI also reopened Pro 200 to new subscribers with a smaller allowance (About ChatGPT Pro tiers).

Verdict: Pay $500 if you already push the old Pro 200 limits in Codex or you want Ultrafast for interactive work with GPT-6 Astra or GPT-6.1 Sol. Stay on Pro 200 if your usage fits inside 10x Plus and speed is not your bottleneck: all three Pro tiers now cost the same per unit of Plus-equivalent usage, so the extra $300 buys scale and Ultrafast, not a discount.

Key Takeaways

• Price and allowance: Pro 500 costs $500 a month and gives 25 times the ChatGPT Plus allowance, OpenAI's highest (OpenAI DevDay recap).
• Ultrafast speed: GPT-6 Astra Ultrafast generates tokens up to 8x faster than Standard mode in Codex, about 300 tokens per second, and up to 6x faster in the API. GPT-6.1 Sol Ultrafast is now available too.
• Ultrafast cost: it uses included subscription limits at 8x the Standard rate and purchased credits at 6x (Codex speed docs).
• Who gets it: among Pro plans, Ultrafast is available only on Pro 500. Buying credits on Pro 100 or Pro 200 does not unlock it. (About ChatGPT Pro tiers)
• Pro 200 changes: eligible existing subscribers keep their previous allowance through October 29, 2026 (About ChatGPT Pro tiers); Sottiaux and The New Stack put the new level at 10x Plus, down from 20x, and The New Stack dates the change to October 30 (The New Stack).
• Against Claude: Anthropic's Max plan tops out at Max 20x for $200 a month, and Claude Code's fast mode is paid from usage credits rather than subscription limits.

ChatGPT Plans Compared: Plus, Pro 100, Pro 200 and Pro 500

The three Pro tiers now scale linearly with price. OpenAI's Codex pricing page lists Pro "Plans at $100, $200, or $500 USD per month" and Plus at $20 (Codex pricing). Tibo Sottiaux, OpenAI's head of product and platform, gave the new multipliers over Plus as Pro 100 5X and Pro 200 10X (Tibo Sottiaux on X), and the DevDay recap puts Pro 500 at 25 times Plus.

Divide price by multiplier and every Pro tier costs $20 per Plus-sized allowance, the same as Plus itself. Under the old Pro 200 terms, 20x for $200 worked out to $10. Sottiaux put the change plainly: "if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan" (Tibo Sottiaux on X; The New Stack).

Codex's pricing page says "Pro plans currently have no five-hour limit," so the Pro allowance is a weekly budget you can spend at your own pace (Codex pricing). Pro 500 also brings the Agents API with computer use to Codex and ChatGPT Work, which OpenAI lists for Pro 500 and Enterprise (OpenAI DevDay recap). For dots, OpenAI's new always-on agents, see our dots explainer.

What Ultrafast Is

Ultrafast is a faster service tier, not a separate model. Introducing it for GPT-6 Astra, OpenAI called it "our premium speed tier for workloads where speed matters most" and said it "offers up to 8× faster token generation (300 tokens per second) in Codex and up to 6x in the API" (OpenAI DevDay recap). Those are token generation speeds, so a task that spends most of its time running tests or tools will not finish 8x sooner.

At launch, GPT-6 Astra was the only model with Ultrafast in ChatGPT Work and Codex. The Codex docs now say "Ultrafast supports GPT-6 Astra and GPT-6.1 Sol" (Codex speed docs); see our GPT-6.1 Sol coverage for that model.

In the API, Ultrafast is broadly available for GPT-6 Astra and GPT-6.1 Sol, for all API users (API Ultrafast guide). Default GPT-6 Astra Ultrafast limits are 500,000 tokens per minute on Build, 1,000,000 on Launch and 5,000,000 on Grow; GPT-6.1 Sol's run from 1,000,000 on Build to 40,000,000 on Grow. GPT-6 Astra Ultrafast supports US data residency and global processing only, while GPT-6.1 Sol Ultrafast also supports EU data residency.

API prices per million tokens at up to 272K input tokens (API pricing):

How Ultrafast Bills Against Your Allowance and Credits

On Pro 500, Ultrafast draws on your plan in order: "Ultrafast uses your included usage first, then your available credits after that allowance runs out" (Codex speed docs). It costs more of each than Standard mode.

These multipliers set how fast each mode spends your allowance and credits, not how fast it runs (Codex pricing). The practical effect: if you ran GPT-6 Astra on Ultrafast all week, Pro 500's 25x allowance would stretch about as far as roughly 3x Plus of Standard Astra usage. At Codex's Standard rates of 250 credits per million input tokens and 1,250 per million output tokens for GPT-6 Astra, the 6x credit multiplier means 1,500 and 7,500 credits once your included usage is gone.

If you sign in to Codex with an API key instead of ChatGPT, "Codex uses API token pricing instead, and ChatGPT credit multipliers don't apply" (Codex speed docs).

Who Can Get Ultrafast

Pro 500 is the only self-serve plan with Ultrafast. OpenAI's Pro tiers page says: "Among Pro plans, Ultrafast is available only on Pro 500," and "At launch, buying credits on Pro 100 or Pro 200 does not unlock Ultrafast." (About ChatGPT Pro tiers). The Codex docs extend that to every other self-serve plan: they "don't have access to Ultrafast at launch, even with purchased credits" (Codex speed docs).

Enterprise has two more conditions. Legacy Enterprise plans that rely on rate limits instead of usage-based billing are not supported, and residency depends on the model: "GPT-6.1 Sol Ultrafast supports inference residency in the United States and Europe (EEA + Switzerland). GPT-6 Astra Ultrafast supports inference residency in the United States only" (Codex speed docs). To subscribe, OpenAI says to "Sign in to ChatGPT and open the pricing page to select a Pro plan" (About ChatGPT Pro tiers).

How to Turn On Ultrafast

• Codex and ChatGPT Work on Pro 500: select Ultrafast in the model picker to run GPT-6 Astra or GPT-6.1 Sol on the faster service tier (About ChatGPT Pro tiers; Codex speed docs). Check what you have left in your usage dashboard or with /status in the Codex CLI (Codex pricing).
• Enterprise: workspace owners enable Ultrafast for selected users or the whole workspace through workspace permissions, and existing per-user spend controls apply (Codex speed docs).
• API: send model: "gpt-6-astra" or model: "gpt-6.1-sol" with service_tier: "ultrafast" on a Responses request. OpenAI recommends a persistent WebSocket connection for agentic apps with many tool calls, because per-request network overhead eats into the latency gain (API Ultrafast guide).

Ultrafast is separate from Fast mode, which you toggle with /fast in the Codex CLI (Codex speed docs). Both draw down the included allowance faster than Standard: OpenAI's Pro tiers page says Fast mode uses it at 2.5 times the Standard rate and GPT-6 Astra Ultrafast at 8 times, and that these are consumption rates, not guarantees of faster performance (About ChatGPT Pro tiers); the Codex pricing docs list GPT-6.1 Sol Ultrafast at the same 8x (Codex pricing).

What Changes for Pro 200 Subscribers

OpenAI's Pro tiers page says "Pro 200 is also available for new subscriptions again," and that new subscriptions not eligible for grandfathering "include a lower usage allowance" (About ChatGPT Pro tiers). The same page sets out who keeps the old allowance. As of October 2026 it states the window as dates: grandfathering applies "if you had an active Pro 200 subscription at any point from September 22, 2026 through 10 a.m. Pacific Time on September 29, 2026," and eligible customers "can use the previous included allowance through October 29, 2026" while their Pro 200 subscription is active, then move to the lower allowance at the same $200 price. Keeping it "does not upgrade your plan or add Ultrafast" (About ChatGPT Pro tiers). OpenAI's own pages do not give the new multiplier; the 10x figure comes from Sottiaux and press briefings.

The New Stack reports the cut to 10x starts October 30, that existing subscribers then get 62,500 usage credits (worth $2,500, OpenAI says) expiring December 31, 2026, and that "subscribers to the $200 Pro plan will also see their GPT-6 Pro allowance in Chat fall from 200 messages per week to 100" (The New Stack). Engadget reports the same 20x-to-10x cut and GPT-6 Pro cap (Engadget). Sottiaux says OpenAI will not reintroduce the five-hour limit (Tibo Sottiaux on X).

Should You Pay $500 or Stay on Pro 200?

Pay for Pro 500 if either of these is true:

• You regularly use more than half of the old Pro 200 allowance. From October 30 that plan gives you 10x Plus. Pro 500's 25x is the only self-serve tier above the old 20x, so it is the one plan that leaves you with more room than you had in September.
• Waiting on GPT-6 Astra or GPT-6.1 Sol is your bottleneck. Ultrafast pays off in interactive loops where you watch the model write: pair programming, debugging, iterating on a UI. It pays off least in long unattended runs where tool calls and tests dominate the clock, and at 8x the included-usage rate it is the fastest way to empty an allowance.

Stay on Pro 200 if you fit inside 10x Plus and do not need Ultrafast on either GPT-6 Astra or GPT-6.1 Sol, or if you work in the background and check results later. Eligible existing subscribers keep their previous allowance through October 29, 2026 (About ChatGPT Pro tiers), so use September's usage dashboard to decide, not the launch-day headline. If you need occasional bursts beyond Pro 200, Plus and Pro users can buy credits without changing plans (Codex pricing).

Teams on ChatGPT Business should compare Business Premium seats instead; Ultrafast is not on Business at launch.

How Pro 500 Compares With Claude Max

Anthropic's Max plan comes in two tiers: "Max 5x: $100 per month" and "Max 20x: $200 per month," with 5 or 20 times the Claude Pro plan's per-session usage (Claude Help Center). Claude sessions reset every five hours and Max plans also have a weekly limit, while ChatGPT Pro plans have no five-hour limit.

The multiples use different baselines, ChatGPT Plus against Claude Pro, so they do not compare directly across vendors; measure your own work on each. What does compare is how speed is sold. Claude Code's fast mode makes Opus "up to 2.5x faster," and for Pro and Max subscribers "fast mode is available via usage credits only and not included in the subscription rate limits" (Claude Code fast mode docs). Pro 500's Ultrafast, by contrast, draws on the included allowance before it touches credits.

At $200 a month, Anthropic still sells a 20x tier while OpenAI now sells 10x at that price and 25x at $500. If raw generation speed on GPT-6 Astra or GPT-6.1 Sol matters more than allowance, Pro 500 is the only self-serve subscription with Ultrafast, and it spends included usage before credits. For how the two harnesses differ beyond price, see our Codex CLI vs Claude Code comparison.

Corrections and Updates

• October 9, 2026: This page said GPT-6 Astra was the only model with Ultrafast in ChatGPT Work and Codex and that GPT-6.1 Sol Ultrafast was coming later. Ultrafast now supports GPT-6.1 Sol too, in the API and on Pro 500, with the same 8x included-usage and 6x credit rates.

Sources

• OpenAI, "DevDay 2026 Recap" (Pro 500, Ultrafast speed, Agents API availability; read through the Internet Archive because openai.com refuses automated requests): https://openai.com/index/devday-2026-recap/
• OpenAI Help Center, "About ChatGPT Pro tiers" (Pro 500 and Pro 200 availability, Pro 200 grandfathering, Ultrafast by tier, how to subscribe; read through the Internet Archive): https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers
• ChatGPT Learn, "Speed" (Ultrafast models, Fast and Ultrafast availability, billing multipliers, Enterprise rules, inference residency): https://learn.chatgpt.com/docs/agent-configuration/speed
• ChatGPT Learn, "Pricing" (Pro plan prices, no five-hour limit on Pro, speed-mode multipliers, credit rates): https://learn.chatgpt.com/docs/pricing
• ChatGPT Learn, "ChatGPT usage limits and spend controls" (Enterprise Ultrafast controls): https://learn.chatgpt.com/docs/enterprise/usage-limits
• OpenAI API docs, "Ultrafast mode" (models, request configuration, rate limits, data residency): https://developers.openai.com/api/docs/guides/ultrafast-mode
• OpenAI API docs, "Pricing" (Standard and Ultrafast token prices): https://developers.openai.com/api/docs/pricing
• Tibo Sottiaux on X, plan multipliers: https://x.com/thsottiaux/status/2104951965184925941
• Tibo Sottiaux on X, Pro 200 usage change: https://x.com/thsottiaux/status/2104823812042940713
• The New Stack, "OpenAI halves $200 plan allowance, launches $500 plan": https://thenewstack.io/openai-halves-200-plan/
• Engadget, "OpenAI adds $500(!) Pro subscription, nerfs its existing $200 tier": https://www.engadget.com/2272106/openai-adds-dollar500-pro-subscription-nerfs-its-existing-dollar200-tier/
• Claude Help Center, "What is the Max plan?": https://support.claude.com/en/articles/11049741-what-is-the-max-plan
• Claude Code docs, "Speed up responses with fast mode": https://code.claude.com/docs/en/fast-mode