Claude vs ChatGPT vs Gemini (2026): Which $20 Plan to Actually Pay For

At $20 a month, Claude Pro, ChatGPT Plus, and Google AI Pro look interchangeable. They are not. The right one is decided by your single heaviest workload, not by which model feels smartest in a one-off test.

The verdict

Buy Claude Pro if you write or code for a living. Opus 4.8 and Sonnet 4.6 hold the longest documents and codebases with the least filler, and Claude Code comes in the box. Buy ChatGPT Plus if you want one tool that does everything passably and a few things excellently: image, voice, video, agents, and the widest plugin ecosystem. Buy Google AI Pro if your work already lives in Gmail, Docs, and Drive, because Gemini 3.1 Pro reads and acts on that context natively and the plan bundles 5TB of storage plus a stack of Google perks.

All three sit within a hair of each other on raw capability. Any of them will draft your email, summarize your PDF, and write a working Python function. So the question is never "which is the smartest model." It is "which one removes the most friction from the thing I do most." Pick that, and the $20 is the easiest decision you will make this quarter.

The one axis that actually separates them

Your heaviest workload decides this, not a benchmark leaderboard.

Here is why. In early 2026 the three frontier labs converged. Opus 4.8, GPT-5.5, and Gemini 3.1 Pro all clear the same bar on reasoning, coding, and writing for normal work. The gap between them on a single prompt is smaller than the gap between you on a good day and a tired day. So a head-to-head "which wrote it better" test mostly measures the prompt, not the model.

What does not converge is the surrounding system. Claude wraps its models in a coding-first product (Claude Code, Projects, long context). ChatGPT wraps its models in the broadest feature surface anyone ships (image generation, Sora video, voice, Codex, agents, custom GPTs). Gemini wraps its models in Google itself (your inbox, your documents, live Search grounding). That wrapper is permanent. The model leapfrogging is monthly and it cancels out.

So ignore the leaderboard screenshots. Ask one question: what is the task you will run a hundred times this month? Optimize for that, and treat the other two as nice-to-haves you can access free.

The numbers, side by side

Current versions and prices, pulled from each vendor's own page:

A note on the model names, because the web is full of stale ones. Anthropic's current flagship is Opus 4.8, not the 4.6 many comparison posts still quote. OpenAI's lineup is the GPT-5.5 family (Instant, Thinking, and Pro), not 5.4. Google's paid flagship is Gemini 3.1 Pro. If an article cites older versions, its prices and limits are probably stale too.

A second number that matters: at the API level, Opus 4.8 runs $5 per million input tokens and $25 per million output; Sonnet 4.6 is $3 and $15; Haiku 4.5 is $1 and $5. You do not pay that on a $20 subscription, but it tells you what heavy use actually costs, which is the math behind every usage cap.

Claude Pro: the builder's and writer's pick

Claude Pro is $20 a month, or $17 if you pay $200 up front for the year. That gets you Opus 4.8 for hard problems, Sonnet 4.6 as the fast default, and Haiku 4.5 for cheap quick tasks, all with Pro usage limits.

What you are really buying is two things the others charge more friction for. First, long context that holds. Drop a 60-page contract or a sprawling codebase in and Claude reasons across all of it without losing the thread halfway down. Second, Claude Code, a terminal coding agent included in Pro, which turns "explain this bug" into "find and fix this bug across these twelve files" and reads like the rest of your code.

The writing is the quiet advantage. Claude produces prose with fewer of the tells that make AI text obvious: less hedging, fewer "in today's landscape" throat-clears, less generic structure. For anyone who ships words, that is hours of editing saved per week.

ChatGPT Plus: the all-rounder

ChatGPT Plus is $20 a month and gives you GPT-5.5 Thinking for advanced reasoning, expanded messages and memory, Projects, tasks, and custom GPTs. Below it sits Go at $8 a month (more GPT-5.5 Instant than free); above it, a Pro tier at $200 a month with GPT-5.5 Pro, unlimited GPT-5.3, and a roughly 1M-token context window for the heaviest users.

Plus wins on breadth. Nothing else at this price does as many things competently in one window: generate an image, talk to it by voice, spin up a short Sora video, run Codex on a repo, kick off Deep Research, or build a custom GPT for a repeatable task. If your week is a grab bag of different jobs rather than one deep one, that range is the value.

The honest limit is that breadth is not depth. On a very long document or a multi-file refactor, a Claude user will usually finish cleaner. And the ecosystem cuts both ways: more features means more surface to learn and more places to get an inconsistent answer.

Google AI Pro: the Google-native pick

Google AI Pro is $19.99 a month and is less an AI subscription than a Google upgrade with Gemini bolted through everything. You get Gemini 3.1 Pro, Deep Research, Gemini 3 Pro inside AI Mode for Search with Deep Search, and Gemini living in Gmail, Docs, and Sheets. The plan also bundles 5TB of cloud storage, $10 a month in Google Cloud credits, YouTube Premium Lite, and Google Home and Health perks.

The math is what makes it interesting. If you already pay Google for storage, a chunk of the $20 is buying things you buy anyway, which makes the AI close to free. And the integration is genuinely different in kind: "summarize the thread from Acme and draft a reply from the attached spec" works because Gemini can actually see your inbox and your Drive, not because you pasted everything in.

The catch is that the magic is proportional to how Google-native you are. Live off Gmail and Docs and it is the smoothest of the three. Live in Slack, Notion, and a non-Google stack and you are paying for integrations you will not touch, and you would get more from Claude or ChatGPT.

There is also a cheaper Google AI Plus tier at $7.99/mo (200GB, lighter Gemini 3.1 Pro access) and a Google AI Ultra tier from $99.99/mo (20TB, up to 20x the limits) for heavy users.

The decision rule

Three questions, in order. Stop at the first yes.

1. 1

   Do you write or code most days?

   Buy Claude Pro. Opus 4.8 and Sonnet 4.6 plus Claude Code and long context win the deep-work case, and the prose needs the least editing.

2. 2

   Does your day live in Gmail, Docs, and Drive?

   Buy Google AI Pro. Native context beats copy-paste, and the storage plus perks make the $20 close to free if you already pay Google.

3. 3

   Is your work a different job every day?

   Buy ChatGPT Plus. The widest feature set is the right tool when there is no single heaviest task to optimize for.

The cost trap to watch: $20 is the floor, not the ceiling. Heavy coding or all-day agent use will hit Pro caps fast, and the real bill is the upgrade tier (Claude Max from $100, ChatGPT Pro at $200, Google AI Ultra from $99.99). Budget for the workload you will actually run, not the demo you tried once.

Still down to two? The pairwise calls go deeper: Claude vs ChatGPT and Gemini vs ChatGPT.

Which is better, ChatGPT, Gemini, or Claude?

None is universally better; at the frontier they are close. The winner is set by your heaviest task: Claude for writing and coding, Gemini for Google-native research, ChatGPT for everything-in-one breadth.

What can Claude do that ChatGPT can't?

Hold very long documents and large codebases coherently while producing prose with less generic AI filler, at the same $20 price. Claude Code, its included terminal coding agent, is also more capable for multi-file work than the equivalent on a Plus plan.

Why are people moving from ChatGPT to Claude?

Mostly writers and developers chasing cleaner output and stronger long-context coding. For general everyday use, most people have no reason to switch, and ChatGPT's breadth still wins.

Is the $20 plan enough, or do I need to upgrade?

For one person doing normal work, $20 is plenty on any of the three. Daily heavy coding, large agent runs, or all-day use will hit the usage caps, which is when the $100-plus tiers (Claude Max, ChatGPT Pro, Google AI Ultra) start to pay for themselves.

Which has the best free tier?

All three are usable free. Gemini's free app is the most generous for everyday questions, Claude's free tier (Sonnet) is the best for writing, and ChatGPT free gives limited GPT-5.5 Instant. Try the one your decision rule points to before paying.

Want the one-page version of this for your own stack? Join the newsletter for the AI tooling breakdowns I send when the models actually change, not every week.