AI agents · OpenClaw · self-hosting · automation

Quick Answer

ChatGPT Pro 500 vs Pro 200 vs Pro 100: Which Plan? (2026)

Published:

The short answer

ChatGPT Pro now has three tiers: Pro 100 ($100/month), Pro 200 ($200) and Pro 500 ($500, new September 29, 2026). Pro 500 is the only one with Ultrafast and has the largest allowance (25x Plus). Pro 200 is back on sale but new sign-ups get a smaller allowance than before. All three include a free dot. Buy Pro 500 if you are limited by usage or latency in Codex; otherwise Pro 100 plus credits is the value pick, and a GPT-6.1 Sol API key is cheaper still for programmatic work.

Plan comparison

Pro 100Pro 200Pro 500
Monthly price (USD)$100$200$500
Included usageBase ProMore than Pro 100; reduced for new non-grandfathered sign-upsHighest; “25x the Plus allowance”
Ultrafast (GPT-6 Astra, up to 300 tok/s)Not included; credits don’t unlock itNot included; credits don’t unlock itIncluded
First dot (always-on agent)IncludedIncludedIncluded
Pro models, Codex, deep research, images, memory, filesYesYesYes
Codex in the cloud, Code Review, Codex Security CloudYesYesYes
GPT-6.1 Sol in ChatGPT Work + CodexYesYesYes
Credits for extra usageYes (Codex, ChatGPT Work, Office add-ins)YesYes; Ultrafast draws from credits after allowance
Annual billingNoNoNo
Sign in with ChatGPT (plan pays in partner apps)YesYesYes

Source: OpenAI Help Center “About ChatGPT Pro tiers,” updated September 29, 2026, and the DevDay recap.

What changed on September 29

  1. Pro 500 launched at $500/month with “our highest usage allowance” and Ultrafast. Sarah Friar told CNBC that when the $200 SKU launched “people thought we’d lost our minds,” and that consumers are “moving more towards consumption” — buying credits rather than seats.
  2. Pro 200 reopened. New subscriptions get a lower allowance; existing ones are grandfathered through October 29, 2026. Only affected users get an email. Keeping the old allowance does not upgrade you or add Ultrafast.
  3. Ultrafast gating. Among Pro plans, only Pro 500. It uses included usage first, then your credit balance. At launch, buying credits on Pro 100/200 does not unlock it. Enterprise also gets it.
  4. Dots included on every Pro plan.
  5. A usage-limit reset was pressed on stage “for everyone in the world” — no official post describes exactly what reset.

Is Ultrafast worth $300 a month?

Ultrafast is the same speed lane OpenAI sells in the API at 6x standard — for GPT-6 Astra, $60 input / $300 output per million tokens versus $10/$50. The Pro 500 premium over Pro 200 is $300/month, so the question is whether you would otherwise buy more than $300 of credits, or whether waiting on Astra costs you more than $300 of time.

  • Yes: you pair-program in Codex on Astra for hours a day and the 8x faster generation changes how you work; or you routinely exhaust Pro 200 and buy credits.
  • No: you mostly chat, research and draft; or your heavy work is batchable. GPT-6.1 Sol at standard speed scores 52 to Astra’s 53 on the Artificial Analysis index and runs in ChatGPT Work and Codex on every Pro plan.

Speed-tier economics across vendors: OpenAI Ultrafast vs Fast mode vs Claude fast mode vs Grok Fast.

Pro vs the API

For programmatic use, Pro is the wrong product. A 30K-input / 5K-output call costs $0.11 on GPT-6.1 Sol and $0.55 on GPT-6 Astra at API list price; $100 of API spend buys roughly 900 Sol calls or 180 Astra calls, with no allowance resets and Batch at 50% off. Pro’s value is the bundle — Codex cloud tasks, deep research, dots, Sign in with ChatGPT paying for partner apps — not raw tokens. If you are choosing between Pro and Anthropic’s Max plans for coding, see cheapest AI coding subscription 2026.

Recommendation

  • Most people: Pro 100. It now includes a dot and GPT-6.1 Sol; add credits in the months you need them.
  • Heavy Codex users already grandfathered on Pro 200: stay through October 29, then re-evaluate against Pro 100 + credits versus Pro 500.
  • Latency-bound Astra users or teams that burn $300+/month in credits: Pro 500.
  • Developers: an API key, with gpt-6.1-sol as the default and Astra on escalation. See GPT-6.1 Sol vs GPT-6 Astra vs Claude Opus 5.5.

Full DevDay context: every DevDay 2026 announcement explained.

Last verified: September 30, 2026. Allowance multiples are OpenAI’s descriptions; OpenAI does not publish absolute message or token counts per plan.

Sources