Claude review 2026: our full test, pricing and limits
Claude is the assistant we recommend to people who write for a living and to people who ship code. Where rivals pile on image generation, video and creative extras, Anthropic has aimed its product at three things: the quality of the prose, the solidity of the reasoning, and the ability to work across very large document sets. We used it daily for several months to find where that strategy pays off, and where it leaves you wanting.
Verdict: the best writing and the best code, in a deliberately narrow product
On long-form writing, on reading a two-hundred-page dossier and on code, Claude is still ahead. It follows instructions to the letter, holds the thread across a long conversation, and invents less than its rivals when it does not know something. In exchange it generates no images at all, visual creation being simply absent from the product, and the Pro subscription runs into usage caps that heavy users hit far sooner than they expect.
What to remember
- The best text and code quality in this comparison, in English and beyond.
- Free with strict caps, then $20 per month for Pro, or $17 billed annually.
- No image or video generation: for visual creation you have to look elsewhere.
What we liked
- +
The most natural prose on the market, with genuine respect for style instructions.
- +
A one-million-token context on recent models, so an entire dossier goes in at once.
- +
Excellent at code, with Claude Code working directly inside a repository.
- +
Projects keep your reference documents alive across conversations.
- +
MCP connectors wire Claude into your own tools without a developer.
What could be better
- –
No image or video generation, where ChatGPT and Gemini both do it natively.
- –
Pro usage caps arrive quickly once you chain long working sessions.
- –
The annual plan is billed in one upfront payment, which stings a small cash flow.
- –
Hosted in the United States, which stays a consideration for sensitive data.
- –
Office integration still runs less deep than Copilot’s inside the Microsoft suite.
Spec sheet
- Vendor
- Anthropic, United States
- Free plan
- Yes, with short caps
- Pro
- $20/month, or $17/month annually ($200 upfront)
- Max
- $100/month (5×) and $200/month (20×)
- Team
- From $25/seat/month
- Context window
- Up to 1 million tokens
- Image generation
- No
- Voice mode
- Yes
- Apps
- Web, iOS, Android, desktop
- Best for
- Long-form writing, document analysis, software development
Prices checked on 3 August 2026 on the vendor’s pricing page, quoted in US dollars excluding tax. The $17 annual rate is billed in a single $200 payment. Anthropic adjusts its usage caps regularly: check the pricing page before you commit, especially if you plan to run Claude Code all day.
Go for it if
- ✓
You write a lot and text quality matters more than a long feature list.
- ✓
You analyse large documents and the context window is your real criterion.
- ✓
You write code and want an assistant that holds up on a real project.
Look elsewhere if
- ✕
You want to generate images or video.
- ✕
You want AI wired natively into your office suite.
- ✕
Your usage is heavy and you never want to meet a cap.
What is Claude exactly?
Claude is the conversational assistant built by Anthropic, a US company founded in 2021 by former OpenAI leaders. You reach it through a browser, mobile apps, a desktop app, and an API for developers.
The positioning is clear from the homepage. Where rivals lead with image generation and voice, Anthropic leads with reasoning, writing and code. The product is narrower and, on its own ground, distinctly more finished.
In 2026 the model family has several tiers: a fast, economical model, a mid-tier model that covers most work, and a top model reserved for the hardest tasks. You pick from a dropdown without wading through jargon, which stays more readable than some rival catalogues.
What are Claude’s key features?
The foundation is conversation with file analysis: documents, PDFs, spreadsheets, images, code. Claude reads everything before answering, and with a context window reaching one million tokens it takes a full dossier without you having to chop it up.
Projects are the most underrated feature in the product. You drop in your reference documents, your style guide, your templates, and every conversation in that project inherits them. No more pasting the same brief into each new session. Cross-conversation memory and Skills, reusable instruction sets loaded on demand, round out the setup.
Artifacts render the output in a separate editable panel, which changes how you work on a long document or a block of code. On the development side, Claude Code works directly inside a repository: it reads files, proposes changes and runs the tests. That is what built the product’s reputation among engineers.
The Pro plan also carries Cowork for office work, a voice mode, web search, and a Claude for Microsoft 365 and Outlook integration. MCP connectors wire Claude into third-party tools, a documentation base or an issue tracker, with no custom build. What is genuinely missing is image generation: for that, look at ChatGPT.
Claude does not win on feature count. It wins because what it does, it does better, and because you spend less time editing afterwards.
How much does Claude really cost?
Anthropic publishes a clear price list with a free tier that genuinely works for occasional use. The thing to understand is that the paid plans differ less by feature than by how much usage they allow.
| Plan | Price | What it covers |
|---|---|---|
| Free | $0 | Occasional use, short caps that reset during the day |
| Pro | $20/month, or $17/month annually | Daily use, Projects, Skills, memory, Claude Code and Cowork |
| Max 5× | $100/month | Five times the Pro allowance, priority at busy hours |
| Max 20× | $200/month | Twenty times the Pro allowance, for all-day heavy use |
| Team | From $25/seat/month | Central administration, one invoice, minimum seat count |
Prices excluding tax, checked on 3 August 2026 on the vendor’s pricing page. The annual rate is billed as a single $200 payment. Usage caps change regularly: check the pricing page before subscribing.
Translated into real use: a writer or consultant handling a few dossiers a day will be perfectly happy on Pro at $20. A developer running Claude Code for most of the day will hit the ceiling before lunch and will need Max, at $100 or $200 a month. That is the main source of unpleasant surprises, and it is better to know it going in.
For a team, weigh $25 per seat here against the $30 per user of Microsoft Copilot, which sits on top of an office licence you already pay for. The maths depends entirely on what your people actually do with the tool.
Explore Claude
Test the free tier on a long piece of writing before paying: that is where the gap with rivals shows most clearly.
Who is Claude really for?
The obvious profile is anyone who writes. Copywriters, consultants, lawyers, communications staff: people whose deliverable is a text gain immediately, because editing gets shorter and the tone lands closer. On demanding prose the gap with rivals is still visible.
The second profile is the developer. Between the quality of the code, the ability to reason across an existing codebase and Claude Code working inside the repository, it is now the reference assistant for many engineering teams.
The third, less obvious, is the analyst: whoever has to swallow a tender document, an annual report or a stack of contracts. A one-million-token context removes the chopping up, and with it the partial answers.
Conversely, if you want a tool that produces visuals, takes dictation or lives inside your inbox, Claude is the wrong pick, and our best AI software comparison will point you somewhere better.
What are the alternatives to Claude?
ChatGPT remains the most versatile assistant: images, voice, web browsing, agents. If you can only pay for one and your usage varies, look there first.
Gemini brings an enormous context window and native integration with Gmail, Docs and Android. For anyone already living in the Google ecosystem, the saved friction is real.
Le Chat by Mistral is the European option: hosted in the EU, built by a French company, very fast to answer. If data location is a contractual requirement where you work, it deserves a serious look.
Finally, Microsoft Copilot plays a different game: its value is its place inside Word, Excel, Outlook and Teams, not raw answer quality. The full picture is in our 2026 comparison.
Our verdict: 4.5/5, the best writer on the market
If your job is writing or coding, the Pro plan pays for itself within a few saved hours.
Frequently asked questions
Is Claude free?
Yes, a free tier exists and gives an accurate feel for the product. It applies short usage caps that reset several times a day, which is fine for occasional use but becomes restrictive as soon as you work with the tool daily. The Pro plan at $20 per month lifts most of those limits, price checked in August 2026.
Is Claude better than ChatGPT?
On long-form writing, reasoning and code, yes, we place it ahead. On visual creation, no: ChatGPT generates images where Claude does not do it at all, and its ecosystem of agents and custom GPTs remains richer. Everywhere else, web search, voice mode, connectors, memory, the gap has largely closed. The right choice therefore depends on what you produce, and our comparison lays out both profiles.
Is my data used to train Claude?
Anthropic hosts its services in the United States. Consumer conversations may be used to improve the models depending on your account settings, which you can turn off. API access and enterprise plans do not train on your data by default. For sensitive material, check the contract that applies and consider the European option.
What is Claude Code?
It is Anthropic’s development tool: instead of pasting code into a chat, you let Claude work directly in your repository. It reads the files, proposes changes and runs the tests. It is included in the paid plans, with far larger allowances on the Max tiers.