OpenAI’s plan lineup has grown by accretion: a free tier, a budget tier, the famous $20 Plus plan, Pro tiers for power users, and business plans sold per seat. The names have also shifted — the plan launched as Team is now called Business. This comparison uses the prices OpenAI lists in October 2026; check the live pricing page before buying, because tiers and allowances change often.
One distinction organises everything: individual plans (Free, Go, Plus, Pro) are for one person’s account, while Business and Enterprise are workspaces an organisation administers, with no training on business data by default.
Free and Go: the on-ramps
The Free plan costs nothing and gives real access to current models with limits on messages, uploads, image generation and advanced features like deep research. For occasional use — a weekly summary, a few drafts — it is genuinely sufficient.
Go, at $8 a month, is the pressure valve: roughly ten times the free tier’s message limits, more uploads and images, and a longer memory of your conversations. It exists for people who bump into free ceilings a few times a week but do not need the full Plus toolkit.
Plus and Pro: the individual ladder
Plus is $20 a month and remains the default recommendation for individuals. It brings priority access to the frontier models, much higher limits, expanded deep research, voice mode, image generation, custom GPTs and agent features. Most professionals who pay for ChatGPT pay for Plus, and most never need more.
Pro is the usage tier, not a capability tier: it starts at $100 a month for about five times Plus’s usage and adds a $200 tier at roughly twenty times, with the longest context windows and the largest deep research allowances. The models are essentially the same; you are buying volume and headroom. Researchers and developers who live in the tool all day are the natural buyers.
Business and Enterprise: buying for a team
Business — the plan formerly called Team — costs $20 per user per month billed annually, or $25 monthly, with a two-seat minimum. It provides a shared workspace, central billing and admin controls, connectors to company tools, and the contractual promise that business data is not used for training by default.
Enterprise is quote-only and annual. It layers on the governance stack large organisations require: single sign-on and SCIM provisioning, enterprise key management, audit and compliance tooling, data residency options and support agreements. Reported per-seat figures cluster far above the Business rate, but the honest price is “ask sales,” and it varies with scale and terms.
Fine print worth knowing
A few details shape the real cost of every tier. In the United States, the Free and Go plans are ad-supported, which is part of how their prices stay low; paid tiers are the ad-free route. Prices also vary by country, taxes are added in many regions, and subscribing through a mobile app store can price differently from the web — buy on the web if the numbers matter to you.
Subscriptions never include the developer API. Tokens bought through OpenAI’s platform are billed separately at per-model rates, and heavy agent or Codex-style usage on consumer plans still counts against plan allowances rather than being truly unlimited. “Unlimited” in this industry always has an asterisk, and the asterisk is a fair-use limit.
Allowances also differ by feature, not just by plan. Deep research runs, image generations, file uploads and voice hours each carry their own counters, which reset on their own schedules. A plan that feels roomy for chat can feel tight for research — match the plan to the feature you actually consume, not the headline.
Finally, plans are repriced and restructured regularly in this market — the Team-to-Business rename and price cut in 2026 is the recent example. Treat any comparison, including this one, as a snapshot, and re-check OpenAI’s pricing page at renewal time.
Choosing, in one paragraph
Try Free until it annoys you weekly; then Go if the annoyance is only limits, Plus if you want the features. Choose Pro when your work is measurably metered in ChatGPT hours. Put a team on Business the moment client data enters chats, because the no-training default and admin controls are the real product. Enterprise is for organisations whose compliance department has already asked the question.
0 Comments