What is Claude AI, and what can it do?
Claude is Anthropic's AI assistant. See how its models differ, what each plan from Free to Enterprise costs, where Claude Code fits, and what it cannot do.
What is Claude?
Claude is an AI assistant built by Anthropic. Technically, it is a large language model (LLM): a program trained on a very large amount of text so it can read and write language. You give it text, files or images. It answers in text. People use it to write and edit, to explain a topic, to summarise long documents, to read screenshots and to write and run code.
You can reach Claude in four ways: a web app at claude.ai, desktop and mobile apps, Claude Code for software work, and the Claude API for your own programs. A free plan exists. Paid plans add stronger models and more usage. The sections below cover the models, the plans, the apps and the limits, and each one links to a deeper guide.
Who makes Claude?
Anthropic makes Claude. Anthropic describes itself as a public benefit corporation, and it says its aim is to secure the benefits of AI and to reduce its risks.
Anthropic's help center says Claude is trained to be helpful, honest and harmless. That does not mean Claude is always right. It writes with confidence even when a fact is out of date. Check anything that matters before you act on it, such as a price or a command you will run on a production server.
Which Claude models are there?
"Claude" is the name of a family of models, not one model. The current family has 4 models. They differ in how hard a problem they can handle and in what they cost. The chart shows the API price of each one, per million tokens. A token is a small piece of text, often part of a word. Input tokens are what you send. Output tokens are what Claude writes back, and they cost more.
The data behind this chart
[
{
"label": "Claude Fable 5.2",
"api_id": "claude-fable-5-2",
"input_usd_per_mtok": 8,
"output_usd_per_mtok": 40,
"notes": "Demanding reasoning and long-horizon agentic work"
},
{
"label": "Claude Opus 5.5",
"api_id": "claude-opus-5-5",
"input_usd_per_mtok": 4,
"output_usd_per_mtok": 20,
"notes": "Long-running agentic coding and knowledge work"
},
{
"label": "Claude Sonnet 5.5",
"api_id": "claude-sonnet-5-5",
"input_usd_per_mtok": 2,
"output_usd_per_mtok": 10,
"notes": "The best combination of speed and intelligence"
},
{
"label": "Claude Haiku 5.5",
"api_id": "claude-haiku-5-5",
"input_usd_per_mtok": "0.10",
"output_usd_per_mtok": "0.50",
"notes": "High-volume, latency-sensitive tasks. Listed as a starting price."
}
]Anthropic's own advice is simple. If you are not sure which model to use, start with Opus 5.5 for most work. Move up to Fable 5.2 for demanding reasoning and long-horizon agentic work, or when Opus at a higher effort level still falls short. Fable is the most expensive tier: $8 per million input tokens and $40 per million output tokens. Sonnet 5.5 is the middle choice when you want speed and intelligence together. Haiku 5.5 is built for high-volume, fast tasks such as classification, extraction and routing, from $0.10 per million input tokens. The guide to choosing between Opus, Sonnet and Haiku walks through real tasks, and there are separate pages on what Fable costs and when it is worth it and on Haiku's price and best uses.
All four current models accept text and images as input and produce text as output. All four handle many languages and can use tools, which means they can call functions that you or an app define. In the API, each of them has a context window of 1M tokens and can write up to 128K tokens in one reply. The context window is how much text the model can hold in view at once: your messages, any files, and its own replies.
Each model ID, such as claude-opus-5-5, is a pinned snapshot. A program that calls that ID gets the same model until you change the ID. Older models such as Fable 5.1 and Fable 5 are still listed in Anthropic's docs. A separate family, Claude Mythos, is available only to organizations that Anthropic verifies through programs such as its Cyber Verification Program, so most people will never see it.
What do the Claude plans cost?
The apps run on a subscription plan. There are two individual paid levels above Pro, and two business plans. Prices below are in US dollars for web subscriptions. Local prices vary by region, so the Pro help article tells you to check claude.ai/upgrade for the price in your country.
The data behind this chart
[
{
"plan": "Free",
"usd_per_month": 0,
"notes": "Sonnet and Haiku. Limited usage every 5 hours and every week."
},
{
"plan": "Pro",
"usd_per_month": 20,
"notes": "Billed monthly. $17/month if billed annually ($200 up front). Adds Opus and Claude Code. Fable runs on usage credits."
},
{
"plan": "Max",
"usd_per_month": 100,
"notes": "5x Pro usage every 5 hours and every week. Fable included, up to half the weekly limit."
},
{
"plan": "Ultra",
"usd_per_month": 200,
"notes": "10x Pro weekly usage. Formerly called Max 20x."
},
{
"plan": "Team standard seat",
"usd_per_month": 25,
"notes": "Per seat, billed monthly. $20 if billed annually. 2 to 150 seats."
},
{
"plan": "Team premium seat",
"usd_per_month": 125,
"notes": "Per seat, billed monthly. $100 if billed annually. 5x the usage of a standard seat."
},
{
"plan": "Enterprise",
"usd_per_month": 20,
"notes": "Per seat, billed annually. Usage is billed separately at API rates."
}
]Free costs nothing. You get chat on the web, desktop and mobile, plus web search, file creation, code execution, memory and Artifacts. The models are Sonnet and Haiku. Opus and Fable are not included. Usage is limited every five hours and every week, and the details are in what the Claude free tier really allows.
Pro costs $20 a month billed monthly, or $17 a month if you pay $200 for a year up front. It adds Opus, Claude Code, Projects, scheduled tasks, Claude in Chrome and the Microsoft 365 integration. Fable is available on Pro, but it is paid from pay-as-you-go usage credits instead of your plan's included usage. The full breakdown is in Claude Pro's price and usage limits.
Max costs $100 a month. It gives five times Pro's usage every five hours and every week, priority access at busy times, and $100 a month in API credits. Fable is included, and it can use up to half of your weekly limit. Read what to check before you upgrade to Max before you pay for it.
Ultra costs $200 a month. This plan used to be called Max 20x, and existing subscribers were moved to Ultra automatically with no change in price. It gives ten times Pro's weekly usage and includes $200 a month in API credits. Verified Ultra accounts have no five-hour limit. The weekly limit still applies, and there is a separate weekly limit for Fable.
Team is for groups of 2 to 150 people. A standard seat costs $25 per member per month billed monthly, or $20 billed annually. A premium seat costs $125 billed monthly, or $100 billed annually, and gives five times the usage of a standard seat. Team adds Claude Code, single sign-on (SSO) and domain capture, central admin and billing, and spend controls. On premium seats Fable is included up to half the weekly limit. On standard seats it runs on usage credits.
Enterprise includes everything in Team, plus audit logs and SCIM (system for cross-domain identity management, the standard that creates and removes user accounts from your identity provider). It has a single seat type, billed annually, and usage is billed on top at API rates. The pricing page lists the seat at $20 per month. The help center says the self-serve route needs at least 20 seats and the sales-assisted route needs at least 50. See how Claude Enterprise pricing works for the full picture.
How do Claude's usage limits work?
Free, Pro, Max, Ultra and seat-based business plans limit your included usage. Usage-based Enterprise plans have no fixed plan or seat-level usage limits; usage is billed by consumption, and admins can set spend limits. A session limit runs on a five-hour window on plans with session limits, but verified Ultra accounts have no five-hour limit. On plans with usage limits, a weekly limit caps your included use across all models. Some plans also have a separate weekly limit for Fable. Paid plans with usage limits can use purchased usage credits to continue beyond their included usage.
The context window also differs by model in the apps. On paid plans, some models support a context window of up to 1M tokens, while others support 500K or 200K. If you hit a limit mid-task, your options when Claude says the limit is reached lists what you can do, and whether usage limits can be reset early answers the question everyone asks next.
Where can you use Claude?
The web app is at claude.ai. The desktop app runs on macOS and on Windows. A Linux desktop app is in beta for Ubuntu and Debian, on x64 and arm64. On phones, Anthropic's download page lists the iOS app, and the Claude Code docs refer to the Claude app for both iOS and Android, so look for it in Google Play on an Android phone.
The Claude API is the route for developers. Your program sends a request with a model ID, and Anthropic bills each token at the API prices in the model chart. The trade-off between paying per token and paying a flat monthly fee is covered in Claude API versus a subscription: which costs less.
What is Claude Code?
Claude Code is Anthropic's agentic coding tool. Agentic means it does more than suggest code. It reads your codebase, edits files, runs commands and works with your development tools. It runs in the terminal, in an IDE (integrated development environment), in the desktop app and in the browser.
There are several ways to get it. Anthropic's docs recommend the native installer for macOS, Linux and WSL (Windows Subsystem for Linux), and that version updates itself. Homebrew and WinGet installs also exist, but they do not update automatically. On Debian, Fedora, RHEL and Alpine you can install it from a package repository with apt, dnf or apk. There is a VS Code extension, and a JetBrains plugin that needs the command-line tool installed separately. The web version at claude.ai/code also works inside the Claude mobile apps.
Most of these need either a Claude subscription or an Anthropic Console account. The desktop app needs a paid subscription. On the pricing page, Claude Code appears from the Pro plan upward. How the two login methods differ in cost and limits is explained in logging in to Claude Code with a subscription or an API key.
Claude Code also runs on a Linux VPS, which stays on when your laptop sleeps. What Claude Code is and how it works is the starting point. Using Claude for server administration covers day-to-day ops work, and building an AI agent with Claude on a VPS goes one step further.
What can't Claude do?
Knowing the limits saves you time. These are the ones that matter most.
- It does not generate photos or illustrations. Anthropic's help center says Claude does not make images the way image-generation tools do. It can build diagrams, charts and interactive visuals in HTML and SVG, which is in beta on the web and desktop apps. It can also view and analyse images you upload.
- Its output is text. All current models take text and images in and give text back.
- The free plan does not include every model. Free covers Sonnet and Haiku only. Opus needs Pro or higher, and Fable needs a paid plan.
- Included usage has a ceiling on the individual plans. Free, Pro, Max and Ultra all have weekly limits, so a large job can stop partway through.
- It can be wrong. Claude can state an outdated price or an invented command flag with full confidence. Run new commands on a test box first.
Two questions come up often enough to have their own pages: whether you can run Claude on your own server, and whether Claude trains on your code and chats.
FAQ
Is Claude free to use?
Yes, there is a Free plan that costs nothing. It includes chat on the web, desktop and mobile, web search, file creation, code execution and memory, with the Sonnet and Haiku models. Usage is limited every five hours and every week. You need a paid plan, starting with Pro, for Opus and Claude Code. Fable also needs a paid plan.
Who makes Claude?
Anthropic makes Claude. Anthropic describes itself as a public benefit corporation. It builds the Claude models and sells access to them through the Claude apps and the Claude API.
Which Claude model should I use?
Anthropic's docs say to start with Claude Opus 5.5 for most work. Use Claude Fable 5.2 for demanding reasoning and long agentic tasks, or when Opus at higher effort still falls short. Pick Claude Sonnet 5.5 for a balance of speed and intelligence, and Claude Haiku 5.5 for high-volume, fast tasks such as classification and routing.
What is the difference between Claude and Claude Code?
Claude is the AI model family and the chat apps built on it. Claude Code is a separate tool that uses Claude to work on software: it reads your codebase, edits files and runs commands in your terminal, IDE, desktop app or browser. Most ways of using Claude Code need a Claude subscription or an Anthropic Console account, and the pricing page lists it from the Pro plan upward.
Can Claude create images?
Claude does not generate photos or illustrations the way image-generation tools do. It can build diagrams, charts and interactive visuals in HTML and SVG, a beta feature on the web and desktop apps, and it can view and analyse images you upload. All current Claude models produce text as their output.