ChatGPT, Claude, and Gemini: Subscription Prices Compared

I'd start with one paid AI subscription, chosen around the work I expect it to do. Paying for three assistants is easy. Finding enough useful work for all three is harder, especially when their free plans already cover occasional use.

The roughly $20 plans are the sensible starting point for regular use. Above that, I want a specific reason to upgrade: a tool I need, a model the cheaper plan doesn't offer, or a usage limit that keeps interrupting work. A more expensive subscription doesn't automatically produce a better answer to an ordinary question.

These are U.S. web subscription prices checked September 30, 2026, before tax. Monthly prices below exclude trials, introductory offers, and student discounts. Annual commitments are identified separately; availability and local pricing can differ.

ChatGPT: a broad toolset, then three Pro tiers

ChatGPT's personal plans now extend from Free through a $500 Pro tier. The first useful distinction is between Go, which adds capacity for everyday use, and Plus, which expands access to reasoning and work tools.

The ChatGPT pricing comparison lists the feature differences. OpenAI's Go announcement gives the $8 U.S. price; its current Go guide explains the scope and possible ads. Plus costs $20 monthly, with no annual billing option.

There is a timing detail worth checking if you're considering Pro 200. It is available to new subscribers again, but the new allowance is lower than the previous one. Eligible existing subscribers retain the older allowance through October 29, 2026, then move to the lower allowance at the same $200 price. All three Pro tiers bill monthly. Astra Ultrafast is exclusive to Pro 500 among those tiers; buying extra credits on Pro 100 or 200 doesn't add it. OpenAI's current Pro guide covers the prices and transition.

I'd consider Plus first for a mix of writing, research, file analysis, and coding. Moving from $20 to $100 adds $80 every month. That needs to solve an actual interruption in the way you work. The jump to $500 needs an even clearer case for the extra capacity or speed.

Claude: Pro for the tools, Max for more usage

Claude has a simpler personal lineup. Pro is the entry point for its paid coding and research tools. Max increases the amount of work you can run through them.

Claude Pro also costs $200 paid annually, about $16.67 per month when spread across the year. The advertised rounded monthly equivalent isn't a monthly payment option. The Claude pricing page lists the plans and included tools.

Max's multipliers compare against Claude Pro's usage allowance. They don't describe model quality, and they don't compare directly with a similarly named Google tier. Claude applies a five-hour session window as well as weekly limits; additional limits can apply to particular models or features. Max is billed monthly. See the Max plan details.

For a developer, Claude Code is a concrete reason to try Pro. I'd give it work from a real project: explain an unfamiliar module, make a bounded change, and run the relevant checks. Keep track of how much correction the result needs. A tool that generates a large patch quickly can still cost more time in review than it saves.

Max becomes easier to justify when you're already satisfied with that workflow and its allowance is the problem. I wouldn't start at $200 just to find out whether I like the product.

Gemini: compare the whole Google bundle

Google sells its personal AI subscriptions under Google AI, with Gemini as the assistant. Storage is part of the price, and several tiers include other subscriptions. That makes a straight comparison with a standalone chat plan incomplete.

Prices and storage are listed in the Google AI plan comparison and Google One plan selector. Google AI Pro also offers $199.99 annual billing, about $16.67 per month. The two Plus options have the same AI tier; the extra $5 buys more storage.

Plus adds AI capacity over free access. Pro expands Gemini access and its integration with Gmail, Docs, and other Google tools, along with Gemini Notebook. The Google One plan selector spells out those differences. If most of your work already sits in Google's apps, using the assistant there may matter more than a small difference in how two chat windows answer the same prompt.

There's also a creative and development side to the bundle: video generation and editing through Gemini and Flow, plus higher Antigravity coding allowances on Ultra. Google's subscription feature overview covers those tools. If video is the reason you're subscribing, compare the media credits and supported editing features separately from the chat allowance.

Pro includes YouTube Premium Lite; Ultra includes YouTube Premium Individual. Those are different subscriptions, and the Ultra benefit isn't a YouTube family plan. Google also lists Home and Health benefits, with eligibility conditions. Check the bundle details against the services you actually use before counting them as savings.

For Gemini usage, Ultra's 5x and 20x labels compare against Google AI Pro, not the free plan. Gemini now meters this through compute allowances, with five-hour and weekly limits. Prompt complexity, model choice, and conversation length can affect consumption. Ultra also provides Deep Think access. The Gemini limits documentation is more useful here than treating “20x” as a promised number of messages.

I'd put Google AI Pro on the shortlist for someone who needs its storage and works heavily in Gmail and Docs. If those extras replace existing bills, the effective cost of adding AI can be lower. If you wouldn't buy them separately, assigning their full retail value to the bundle makes the deal look better than it is.

A subscription isn't a general API budget

ChatGPT and Claude subscriptions don't include general API usage. OpenAI explains this in the Plus billing details; Anthropic separates Claude subscriptions from API and Console billing.

An included coding agent and an application you build around a provider's API can therefore have different bills. Before paying for an upgrade to support a script or service, check which account and billing path that software actually uses.

Google AI Pro and Ultra include eligible Google Cloud developer credits. These have their own conditions and are separate from media-generation credits. Check the developer benefits before budgeting for API use; the subscription doesn't make that usage unlimited.

Buying for a team changes the comparison

For work accounts, start with account ownership and administration. Someone needs to manage access when a person joins or leaves, and the organization needs to know which terms cover its data. A larger personal allowance doesn't resolve those requirements.

Both require at least two seats. Premium mainly adds capacity over Standard. The ChatGPT Business overview and Claude Team guide cover the seat options, shared administration, and billing.

Google Workspace is a different bundle again: business email, collaboration, storage, and AI. Its regular U.S. Business prices are $7 for Starter, $14 for Standard, and $22 for Plus per user per month with a one-year commitment. Those plans are billed monthly under that commitment. Standard broadens the AI integration beyond Starter's Gmail access; Plus adds storage and administrative features such as Vault. See Google Workspace pricing.

These are starting points for a team comparison, not interchangeable feature sets. For an enterprise purchase, get the full quote, included usage, retention terms, and required controls in writing. A published seat price may be only part of the bill.

What I'd pay for first

For occasional questions and light drafting, I'd stay on the free plans until something specific gets in the way. Google AI Plus and ChatGPT Go provide cheaper steps up when everyday capacity is the main issue.

For regular work, I'd trial one of the roughly $20 plans for a month. ChatGPT Plus is a useful starting point for varied tasks across its tools. Claude Pro deserves a close look if Claude Code is central to the work. Google AI Pro is particularly interesting when its app integration and storage replace things you already pay for. Those are workflow choices, not a claim that one model wins every task.

I'd use the same small set of tasks during that month: a piece of writing that needs revision, a document that needs careful reading, and a coding or analysis task with an answer I can check. Record useful results, corrections, and the times a limit stops the work. Keep a free account with another provider for a second opinion.

Upgrade to $100 or more when that record shows a recurring problem the higher tier actually solves. Try the higher tier for a month before taking on a longer commitment where one is offered. If the cheaper plan already gets the work done, keep the difference.