Skip to content
Neural Minute

Review · AI chatbots and companions

Claude AI Review 2026: Anthropic’s Chatbot Tested

By Maxime Aulne, Founder and Editor

Claude is free to start, Pro costs $20 a month, Max $100 and Ultra $200. What each plan opens, where the limits bite, and who should pay.

Minute Verdict

Claude is Anthropic’s assistant, free at $0 and $20 a month for Pro, with Max at $100 and Ultra at $200. The subscription covers chat on web, desktop and mobile, Artifacts, Projects, Research, Skills, Claude Code and Claude in Chrome, and all of it draws on one usage pool.

Buy Pro if you want one payment to cover a chat window, long documents and a coding agent in the terminal, since Claude Code is included on every paid plan. Skip it if you need a number before you pay: Claude sells a five-hour window and a weekly allowance, and never a message count.

Best for: a writer or developer who works in one assistant all day and wants the coding tool in the same bill.

Not for: a buyer who needs a predictable monthly volume, or anyone who only wants image generation.

From: $20/mo billed monthly, checked November 15, 2026

What Claude does

Claude is the assistant built by Anthropic, sold as four individual plans and two business offers. The pitch is breadth: the same subscription opens a chat window on web, iOS, Android and desktop, a set of document tools called Artifacts, Design, Slides, Docs and Sheets, a research mode, a library of reusable Skills, and Claude Code for work in a terminal or a repository.

The models carry the plans. Sonnet 5.5 and Haiku 5.5 are on every plan including Free, Opus 5.5 starts at Pro, and Fable 5.2, the top of the range, is included on Max and Ultra up to half of the weekly limit and reachable on Pro only through usage credits. All four current models take text and images as input and publish a 1M token context window, which is the practical reason people put whole contracts and codebases into a single conversation.

The product list also explains what Claude is not. There is no image generator in the lineup: the visual tools are Artifacts, Claude Design, Slides and Docs, which assemble documents, diagrams and decks. Claude Code writes and edits code from a terminal and draws on the same pool as the chat window, so a coding session spends the same budget. Claude Design, the layout tool sitting in that list, has its own Claude Design review.

On this page

What you get for the money

Plan Price What is included The limit that matters
Free $0 Sonnet and Haiku, web search, files, code execution, memory, up to 5 projects Limited usage every 5 hours and every week, no Claude Code, no Opus
Pro $20/mo, or $17/mo with $200 a year up front Everything in Free, more usage, Opus, Claude Code, Projects, Research, Skills, Claude in Chrome Fable only through usage credits, usage still capped per session and per week
Max $100/mo 5x Pro usage, Fable up to half the weekly limit, $100 a month in API credits, priority at peak Billed monthly only, one seat
Ultra $200/mo 10x Pro weekly usage, no 5-hour limit on verified accounts, $200 a month in API credits Billed monthly only, one seat
Team Standard $25/mo, or $20 billed annually Every Claude feature with more usage than Pro No invoicing or ACH below Enterprise
Team Premium $125/mo, or $100 billed annually 5x the usage of a Standard seat The expensive seat is the only one with real volume
Enterprise $20/seat/mo billed annually, 20 seats minimum Team features plus SCIM, audit logs, spend limits, IP allowlisting Usage billed at API rates on top of the seat

Pro is the only plan that changes what Claude is. Below it you have a restricted assistant without the terminal tool and without Opus. At $20 you get the full product for a month, and twelve monthly payments come to $240 against $200 for the annual view, which is the one place where Anthropic publishes a real discount.

Above Pro, every dollar buys volume and nothing else. Max multiplies the Pro allowance by five, Ultra multiplies the weekly allowance by ten and removes the five-hour window for verified accounts, and both return part of the money as API credits. For a single reader that is a fair trade only if the limits are hit every week.

Where it falls short

The first problem is that no limit is quantified. A five-hour window and a weekly allowance exist on every paid plan, and the factors that drain them are named: the model, the length of the conversation, the size of the files, the tools you switch on. None of that converts into prompts. A buyer choosing between $20 and $100 has to pay for a month of each to find out.

The second is that the best model sits behind a meter. Fable is included on Max and Ultra only up to half the weekly allowance, and on Pro it is billed through usage credits at API rates, so the plan most people will buy reaches the flagship model by paying again for it.

The third is the shared pool. Chat on the web, the desktop app, the phone and Claude Code all count against the same budget, so a long coding session in the terminal spends the allowance a reader was saving for writing. Nothing warns you before the two collide, beyond the usage table in settings.

The fourth is the training line. Individual plans list model training as opt-out, which means a setting to change rather than a default. Only Team and Enterprise state no training on your content by default, and that difference is worth a compliance conversation before a company rolls Claude out.

The last is money out. Cancelling stops the next charge and leaves the plan running until the period ends, but payments are generally non-refundable, and no trial is published on any plan. The free tier is the trial.

Claude alternatives

Three assistants sell against Claude, and each is priced differently enough to matter.

  • ChatGPT, from $8 a month for Go and $20 for Plus, the closest match on price and the one that also bundles a coding tool in the paid plans: ChatGPT review.
  • Perplexity, $17 a month billed yearly, built around answers with sources rather than long documents, so it fits research more than drafting: Perplexity review.
  • DeepSeek, free as a chat and from $0.003 per million input tokens on the API, which is the cheap route if you build rather than type: DeepSeek review.

Claude is the only one of the three that puts a document suite and a terminal agent inside the consumer plan, which is the argument for its $20.

Who it is for: Claude

Claude suits people whose work is text and code in the same day. A writer who feeds a 60 page report into the chat can ask for a summary, a set of slides and a critique of the argument from one subscription at $20. A developer who wants the terminal agent without buying a second subscription gets Claude Code in the same bill. The AI chatbots category shows how that compares with the rest of the field.

It does not suit someone buying volume by the message. If your month is mostly short questions, the Free plan covers a surprising amount, and paying $20 buys a ceiling you may never touch. It also does not suit anyone who needs image generation, or a company that needs invoicing and audit logs without going to Enterprise, since none of the individual plans and neither Team seat settles those.

How we checked

This is a research review: we have not used Claude, so there is no hands-on score and no output sample. Every price on this page was read on November 15, 2026 on claude.com, in both billing views, and the per-model API rates come from the same place. The session and weekly limit rules come from Anthropic’s usage documentation and the model lineup from its platform documentation, both read the same day. What a paid month actually buys in messages is not verified, because no plan publishes a count, and a hands-on review will replace this one once we run a paid Pro month.

What to do next

Open the Free plan and spend a week with it on real work, not on test prompts. If you hit the five-hour limit twice in that week, pay $20 for Pro, which is where Claude Code, Opus and Projects appear. Take the annual view at $200 only after a month of Pro, since the $40 saving locks the money for a year. Consider Max at $100 only if you already know you hit the weekly ceiling.

FAQ

Is Claude free? Yes. The Free plan costs $0 and carries limited usage every five hours and every week, with Sonnet and Haiku, web search, files, code execution, memory and up to five projects. Claude Code, Opus and Fable are not in it.

How much does Claude cost? Pro is $20 a month billed monthly, or $17 a month with $200 paid up front for the year. Max is $100 and Ultra $200. Team seats are $25 and $125 billed monthly, or $20 and $100 billed annually. Enterprise is $20 a seat a month billed annually, 20 seats minimum, plus usage at API rates.

Is Claude worth it next to ChatGPT? Both entry plans sit at $20 a month, so price does not decide it. Claude puts Claude Code, Projects and Research inside that $20, and ChatGPT sells its own coding tool inside Plus and Pro. Compare the limits rather than the feature lists, since neither company publishes a message count.

Can Claude generate images? No image generator appears among the Claude products. The visual tools that are included are Artifacts, Claude Design, Slides and Docs, which lay out documents, diagrams and presentations rather than producing pictures from a prompt.

Does Claude have a free trial? No trial is published. The Free plan is permanent rather than timed, and every paid plan can be cancelled before the next billing date, but payments are generally non-refundable once taken.

Is Claude safe to use at work? Anthropic publishes consumer terms and a privacy policy, and the training line differs by plan. Individual plans are opt-out, while Team and Enterprise state no training on your content by default. Companies with a compliance review should start on Team.

Sources

Sources

Update log

November 15, 2026: first publication, research review, all prices and API rates read on claude.com on the same day in both billing views.

Newsletter

One email when a review lands.

That is when the promo codes are still fresh, before the prices move.

Nothing else. Unsubscribe in one click.