Claude

Claude is Anthropic's AI assistant, and in my opinion it's the one to beat for writing, long documents and code. You can use it for free on the web, on your phone or in the desktop app, with limits. If you use it every day, Pro costs $20 a month before tax ($17 if you pay the year up front), adds the Opus model and Claude Code, and is the plan I'd recommend to most people. Max ($100 a month) and Ultra ($200) are for people who work in Claude all day.

That's the short version. Here's what Claude is actually good at, what the free plan includes, and when paying makes sense.

What Claude is, in plain English

Claude is a chat assistant, like the ones you've probably tried already: you type (or talk), it answers. You can ask it to draft an email, rewrite a paragraph so it sounds less like fine print, explain a concept, compare two options or turn your messy notes into a plan. You can also drop files into the conversation, a PDF, a spreadsheet, a screenshot, and ask questions about them.

On top of the chat, Claude can search the web, remember things across conversations, create documents and other files for you by running code behind the scenes, and build small interactive things called Artifacts that open next to the chat. Projects let you keep related chats and reference files together, so you don't have to paste the same background every time.

Sign-up is quick: a Google account or an email address, and you're in. It works in the browser, in desktop apps for Mac and Windows (Linux is in beta), and on iPhone and Android.

Where it shines

Long texts. This is where Claude earns its reputation. The plan comparison lists a context window of "up to 1M" tokens, varying by model. Sorry, a bit of jargon: tokens are chunks of text, and a bigger window means Claude can keep a much larger document in mind at once. In practice, that's the difference between "summarize this report" and "read these three reports and tell me where they contradict each other."

Writing and rewriting. Drafts, rewrites, tone changes, turning a list of bullet points into an email you'd actually send. The trick with any assistant is to be specific: "shorter, warmer, no buzzwords" gets you further than "make it better." You'll still want to read everything before you hit send, obviously.

Code. For developers, the big draw has a name: Claude Code.

Claude Code: the part I use every day

Claude Code is Anthropic's coding agent. Instead of pasting code into a chat window, you give it a task ("fix this bug," "add a test," "update these files") and it reads your project, proposes changes and makes them, with you reviewing along the way. It runs in the terminal on macOS, Linux and Windows, and the Claude Code page also lists the web, iOS, Android, GitHub, VS Code, JetBrains and Slack.

This one I can talk about from experience. Claude Code is installed on the server where I manage my websites, and I use it every day to get real work done there. What I'd tell you before you jump in: it's included in every paid plan, and it shares the same usage limits as your chats. A long coding afternoon and your regular chatting come out of the same pool, which is why heavy coders are the ones most likely to need Max. You can also use Claude inside other coding tools, but it isn't necessarily cheaper: in Cursor, Claude is only available from the $30 Premium plan up, and billed on top at API prices.

Free vs. paid: where the line is

The free plan is more than a teaser. According to Anthropic's plan comparison, on Free you get chat on web, desktop and mobile, web search, memory, voice mode, file creation with code execution, Artifacts, Connectors to your apps, up to 5 Projects, and Claude Design, Slides and Docs. You get the Sonnet and Haiku models.

Connectors are also the back door for the one thing Claude doesn't do on its own, images and video: with Higgsfield MCP it generates them through your Higgsfield account. That part takes a paid Higgsfield plan, and every image or clip spends Higgsfield credits.

What you don't get on Free:

• Opus, the model Anthropic positions for agentic coding and enterprise work.
• Fable 5.1, the most capable model, which is only on paid plans and with conditions: here's who gets it.
• Claude Code.
• Research, Claude in Chrome and Claude for Microsoft 365.
• Usage credits to keep going once you hit the limit.

That last one, usage credits, matters more than it looks. Anthropic doesn't publish a fixed message count, because it depends on how long your conversations are, what you attach and which model you use.

My rule of thumb: the day you catch yourself waiting for the reset, or wishing you had Opus for something tricky, that's your sign to pay.

Which plan I'd pick

Max costs 5 times monthly Pro for 5 times the usage, so what you're really buying is not getting cut off mid-afternoon. Ultra, the plan that used to be called Max 20x, costs 10 times monthly Pro for 10 times Pro's weekly usage, and on a verified account it drops the five-hour limit altogether: it's for long jobs that keep stopping halfway on Max. Anthropic says it verifies Ultra accounts automatically and applies a temporary limit until it does. Both plans are billed monthly only, and both come with monthly API credits for the Claude Platform: $100 on Max, $200 on Ultra.

One more thing before you pay: Anthropic says it doesn't offer standing discounts on its paid plans, so waiting for a sale isn't much of a strategy. Paying Pro annually is the one reliable saving: $200 instead of $240. For Team seats, Enterprise and the API rates model by model (the new Haiku 5.5 starts at $0.10 per million input tokens), I've put everything in the Claude pricing guide.

Privacy, briefly

On Free, Pro, Max and Ultra, whether your chats are used to train Anthropic's models is an opt-out setting: it's on unless you turn it off, so check it in your settings if that matters to you. On the Team and Enterprise plans, the pricing page says your content isn't used for training by default. If you handle client data, that's a good reason to look at Team instead of individual Pro accounts.

Claude, ChatGPT or something else?

If you're choosing between Claude and ChatGPT, honestly, don't decide from a review, mine included. Both have free plans: use each one for a week on your real work and pay for the one you keep opening. ChatGPT is the better pick if you also want images; Claude, if your days are full of long documents or code, and especially if you'd use Claude Code, which comes with the $20 plan.

If what you mostly do is look things up, Perplexity is built around search with sources, and if you'd like several chatbots under one roof, Claude included, Poe does that. And if you'd rather hand over a whole task (research, a slide deck, a website) than chat your way through it, that's what Manus is built for. And if the thing you want is a small web app but you'd rather not touch the code yourself, Lovable and v0 are app builders: you describe the app in plain English and they write it and run it for you. Lovable's free plan gives you 5 credits a day; v0 starts you off with a one-time $5 credit.

Me? Claude Code is open on my server every single day, so you can guess which side I'm on. 🙂

Written by Miguel Ángel, with help from an LLM to verify facts.