Claude Pro vs ChatGPT Plus: what each $20 plan actually includes

Twenty dollars buys you very different things. Claude Pro spends it on depth: stronger models for reading, reasoning and writing, a research agent, and a coding agent. ChatGPT Plus spends it on range: images, voice, a browsing agent and a coding agent, in one app.

Sources: claude.com/pricing, learn.chatgpt.com and OpenAI's help centre, checked 22 October 2026.

A caution on ChatGPT models. OpenAI's own pages disagree about which GPT-6 models reach the ordinary Chat window on Plus and which are limited to its Work and Codex surfaces. Until that settles, look at the model picker on your account rather than any comparison table, this one included.

Our pick for anyone whose work is mostly reading, analysing and writing: Claude Pro. Opus 5.5 is on it, and Opus currently tops Artificial Analysis's Intelligence Index at 58. For someone who wants one subscription to do images, voice, quick web tasks and a bit of code, Plus is the better twenty dollars.

Neither company sells you a number of messages. Both meter usage over five hours, and both let you buy your way past the limit. The details differ.

Claude Pro

• Usage resets on a rolling five-hour session window, and paid plans add weekly limits on top (claude.com/pricing FAQ).
• Pro gives "at least 5x more usage per 5-hour session than Free". Anthropic publishes "no fixed message count", because length, attachments and tools all change the cost of a message.
• Research sessions "can use up your limits faster" (Claude help centre), and so does code execution, which Anthropic says uses more plan usage than normal chat.
• When you hit the limit you can turn on usage credits and keep going. Fable 5.1 is only available on Pro this way.

ChatGPT Plus

OpenAI publishes ranges for its Work and Codex surfaces, per five hours, for Plus and Business Standard (learn.chatgpt.com, checked 22 October 2026):

Read the width of those ranges as the real information: a short question and a long agent task are different amounts of work, and the limit tracks the work. On top, ChatGPT credits extend usage, and deep research shows its remaining allowance in the product rather than on a public page.

In practice: on Claude, long documents and Research runs drain the window. On ChatGPT, choosing the most capable model does.

Whichever plan you pick, the answers need the same checking. Count connects to Claude through its MCP server and labels each claim in an answer, so that check takes minutes.

Choose Claude Pro and you gain

• Opus as well as Sonnet. Free users only get Sonnet and Haiku. For a fuller picture of the range, see Claude Sonnet vs Opus.
• Research, which isn't on Claude's Free plan at all.
• Long-document handling. Up to 20 files per chat, 500MB per file and PDFs up to 1,000 pages (Claude help centre).
• Claude Code, Claude in Chrome and Microsoft 365, scheduled tasks, and Claude Dashboards (beta).

Choose Claude Pro and you give up

• Image generation. Claude doesn't offer it on any plan.
• A cheaper step. There's nothing between Claude Free and Pro; ChatGPT has Go at $8.
• Room for big data files in code. Claude's code sandbox handles files up to 30MB, where ChatGPT accepts spreadsheets of roughly 50MB.

Choose ChatGPT Plus and you gain

• Image generation and voice in the same app.
• ChatGPT Work for longer, multi-step tasks, with a cloud browser for web workflows (it replaced agent mode, which OpenAI has retired), and Codex with cloud code review.
• Larger uploads: a 512MB hard limit per file, with spreadsheets around 50MB depending on row size (OpenAI file uploads FAQ).

Choose ChatGPT Plus and you give up

• Creating new custom GPTs. OpenAI says new GPT creation isn't available on personal accounts, Plus and Pro included; existing GPTs still work.
• Certainty about your model. See the caution above.

If you can spare $40, running both isn't extravagant. Two models from different labs agreeing on a claim is weak evidence; two models disagreeing is a strong signal to check. We cover how they differ on accuracy in Claude vs ChatGPT.

Both companies run a ladder above $20. Climb it for one reason only: you're hitting limits often enough that it costs you work.

Sources: Anthropic's Max plan article and pricing page; learn.chatgpt.com and OpenAI's agent help article.

Go to Claude Max if you live in Claude Code or run Research several times a day, or if you want Fable without paying credits. The bundled API credits make it better value if you also build on Anthropic's API.

Go to ChatGPT Pro if you work in Codex or ChatGPT Work for hours at a stretch and keep hitting the five-hour limit, which OpenAI says Pro plans currently don't have. The $500 tier is for people running agents most of the day; few will need it.

For teams, both have a business plan at $20 per seat billed annually and $25 monthly (Claude Team standard seat, ChatGPT Business with two or more users). OpenAI says Business data isn't used for training by default.

What the upgrade from free buys, on both sides, is access to stronger models and research agents. That raises the ceiling. It doesn't change the floor: both still produce confident claims with nothing under them.

The research features are where this bites, because the output looks most finished. When Columbia's Tow Center tested AI search tools on where 200 news excerpts came from, ChatGPT Search got 134 wrong and declined none (Tow Center, March 2025). Claude wasn't in that study, and no independent study of Claude's Research citations has been published, so the absence of a bad result is not a good one. More in Claude vs ChatGPT for research.

A first-week check, whichever you buy

1. Open the model picker and write down which model you're on. You'd be surprised how often it isn't the one you assumed.
2. Give it one task you've already done, with a known answer: a report, a reconciliation, a summary you checked.
3. Open three of its citations and read the sentence it relied on. Does the page say that?
4. Ask: "What did you assume that isn't in my files?" Count the assumptions it admits to.
5. Keep a tally of limit hits for the week. That tally, not a feature list, tells you whether to upgrade.

Step 3 is the slow one. Count does it from inside Claude or ChatGPT: ask it to audit the answer and it returns every claim with its reasoning chain, marked as a citation, an assumption, model knowledge or a hallucination.

Is Claude Pro worth it over ChatGPT Plus?

Which has higher usage limits, Claude Pro or ChatGPT Plus?

Does Claude Pro include Opus?

Is there a cheaper way to get Claude Pro or ChatGPT Plus?

Can Claude Pro generate images like ChatGPT Plus?