Start on the free tier and pay only when a limit stops you. In the US, OpenAI lists four individual tiers: Free at $0, Go at $8 a month, Plus at $20 and Pro at $100, $200 or $500 (OpenAI’s help centre and Go announcement, read 29 Sep 2026). Its pricing page shows euros to a Dutch visitor: Go EUR 8, Plus EUR 23 and Pro from EUR 103. The first three differ mostly in how much you get, not in what you can do; the step that changes capability is Plus, which adds the advanced reasoning models (the pricing page also lists projects, scheduled tasks and custom GPTs against Plus, while OpenAI’s Go help article says Go includes projects, tasks and custom GPTs, so check the plan you are buying). Earlier versions of this review described GPT-4o; the page no longer lists it.
What each tier adds
| Tier | US price (per month) | Euro price shown in the Netherlands | What the page says it adds |
|---|---|---|---|
| Free | $0 | 0 | Unlimited text chats “as long as there is no abuse”, limited uploads, image creation, voice, deep research and memory |
| Go | $8 | 8 | More messages with tools, uploads, image generation, voice chats and memory; the page notes this plan can show ads |
| Plus | $20 | 23 | Advanced reasoning models, extended limits, more accurate image generation, extended deep research and memory, projects, scheduled tasks, custom GPTs, extended Codex use, early access to features |
| Pro | $100 (usage level 1), $200 (level 2), $500 (level 3, adds Ultrafast) | from 103 | Your choice of three usage levels, Pro-level reasoning, longer Codex and ChatGPT Work sessions, unlimited faster image creation, maximum deep research and memory |
The euro figures are regional: they are what the pricing page showed to a Dutch visitor and probably include VAT, so they are not a conversion of the dollar prices. The dollar prices come from OpenAI’s help centre (Plus $20 a month, billed monthly only; Pro plans $100, $200 and $500) and its ChatGPT Go announcement ($8 in the US, localised in some markets). OpenAI says all purchases are in USD except in a limited set of countries with local-currency billing. Business and Enterprise are priced separately and are not covered.
Context window and input size
The comparison table gives a total context window for the fast “Instant” models of 27K tokens on Free, 54K on Go and Plus, and 128K on Pro; for reasoning models it says 256K on Go and Plus and 400K on Pro. It translates the maximum input into pages of text: about 12 pages on Free, about 40 on Go and Plus, about 250 on Pro for Instant models, and about 320 (Go, Plus) or 680 (Pro) for reasoning. So long-document work is one reason to move up a tier.
Who it suits, who should skip
- Stay on Free if you mostly ask questions and draft short text: the page says text chats are unlimited.
- Go suits people who hit the free upload, image or voice limits but do not need the reasoning models, and can tolerate ads.
- Plus suits people who want the stronger reasoning models and extended limits; the projects, scheduled tasks and custom GPTs listed against it on the pricing page are also said to be on Go in OpenAI’s help article.
- Pro is for heavy Codex and research use; the lower tiers cover ordinary writing and analysis.
- Look elsewhere for cited research (Perplexity) or long documents where you prefer Claude or Gemini; we have not compared output quality.
Privacy
Every individual tier on the table shows “opt-out available” for using your content to train models. Check the setting after signing up.
What this page cannot tell you
Answer quality, hallucination rates and the exact message caps are not on the pricing page (“limits apply”), and we have not assessed them. Image generation moved from DALL-E to newer image models; see Midjourney vs DALL-E 3 and the image generator guide. Two-assistant comparison: ChatGPT vs Claude.
Method
Researched from OpenAI’s pricing page, read 29 Sep 2026 in a browser session in the Netherlands (euro prices, Dutch-language page translated by us), and from OpenAI’s help centre and Go announcement for the US dollar prices. Nothing here comes from using the product.