A deep read: the full picture, with the receipts.
Most people asking this question do not need the answer they are usually given. The honest version is that 4 of the major providers charge within a few dollars of the same price for their mainstream plan, they are all good, and the deciding factor is almost never model quality: it is which limits you hit and which apps you already live in.
Prices below were verified in June 2026 and this market moves fast. Treat the exact figures as a snapshot and check the provider before you subscribe; the shape of the decision changes far more slowly than the numbers.
Start with free, because free is now genuinely good#
There are 5 free tiers among the major providers, and they are no longer demos.
If a free tier covers you, that is the correct answer and no amount of comparison changes it. The reason to pay is almost always one of two things: you keep hitting a cap, or you want a specific feature that sits behind the paywall.
The mainstream tier, where nearly everyone lands#
OpenAI ChatGPT starts at $8 (Go), Anthropic Claude starts at $20 (Pro), Google Gemini starts at $4.99 (Google AI Plus), Perplexity starts at $20 (Pro), xAI Grok starts at $30 (SuperGrok).
Most providers cluster their main consumer plan around the same price:
At this tier you are not really choosing a model, you are choosing a set of limits and an ecosystem. Ask which of these is true for you:
- You live in Google apps. Gemini's paid tiers put AI inside Gmail and Docs and bundle storage: that bundle is worth more than a marginal quality difference.
- You write code. Claude's Pro plan bundles Claude Code, which is a different kind of thing from a chat window.
- You want breadth of features. ChatGPT Plus carries the widest feature set (video, agents, deep research) at the mainstream price.
- You want sourced answers. Perplexity is an answer engine with citations first and a chatbot second.
- You live on X. Grok's edge is real-time access to it.
When the expensive tiers make sense#
The $100 and $200 plans are not better models so much as more of them: headroom, priority, and higher caps. They are worth it only if you are already hitting limits on a mainstream plan regularly enough to be annoyed. If you are not sure, you are not.
The question almost nobody asks#
If your usage is spiky (heavy some weeks, nothing for a fortnight), a subscription may be the wrong shape entirely, and paying per-token through an API can cost less. That is a genuinely different calculation and it depends on your volume, so work it out rather than guessing: subscription vs API.
To filter every plan here by provider, price and limits, use the AI subscription comparison.
Frequently asked questions
What is the cheapest way into a paid AI assistant?
As of June 2026, Google AI Plus at $4.99 a month: cheapest paid major-provider tier, suited to light paid users wanting Gmail/Docs AI. What a cheap tier adds varies by provider. Some raise your usage, some also move you up a model tier, so read the plan rather than assuming.
Is the $20 tier enough for most people?
Yes, for most people, most of the time. The mainstream plans are where providers put their best consumer feature sets. Step up only when you have hit the caps often enough to know it, not in anticipation.
Should I pay for more than one?
Rarely worth it. The overlap between these plans is large and the differences are mostly ecosystem and limits. If you find yourself wanting two, that is usually a sign one of them is the wrong fit rather than that you need both.
Do these prices include tax?
No, the figures here are the headline monthly prices the providers publish. Local tax and currency conversion are added at checkout and vary by country.
Sources
- Anthropic Claude pricingclaude.com
Discussion