What Claude is and why teams choose it, the full model lineup, every plan and price, API token economics, downloads for each platform, and honest comparisons with other frontier AI - all on one page.
Claude is a family of large language models built by Anthropic, and the suite of products around them. Chat with it in a browser or app, put it to work in your terminal, spreadsheets, and slides, or build your own software on the API.
Claude answers questions, writes and edits, analyzes documents and images, codes, searches the web, and runs multi-step "agentic" tasks - available on web, iOS, Android, and desktop.
High-quality writing and coding, very large context windows (up to 1M tokens), precise instruction-following, honest and careful answers, and Anthropic's safety-first research approach.
claude.ai/new drops you straight into a fresh conversation - attach files, enable web search, create artifacts, and organize work into Projects.
Anthropic's lineup runs in tiers: Mythos-class (a new top tier above Opus), Opus (flagship), Sonnet (balanced default), and Haiku (fast and economical).
| Model | Tier & role | Context / max out | API price · in / out per 1M tokens |
|---|---|---|---|
| Claude Fable 5Top tierclaude-fable-5 | First Mythos-class model - the most intelligent generally available Claude, with additional safety measures for dual-use capabilities. Built for the hardest reasoning and long-horizon agent work. | 1M / 128K | $10 / $50 |
| Claude Mythos 5approved orgs only | The same underlying model as Fable 5, offered without those additional safeguards to approved organizations only. | 1M / 128K | $10 / $50 |
| Claude Opus 4.8claude-opus-4-8 | Flagship Opus (May 2026). Adaptive thinking, effort controls, and an optional Fast Mode at 2× standard pricing for up to 2.5× speed. | 1M / 128K | $5 / $25 |
| Claude Sonnet 5Newdefault on Free & Pro | Launched June 30, 2026 - the most agentic Sonnet yet, performing close to Opus at a fraction of the cost. | 1M | $2 / $10* |
| Claude Sonnet 4.6claude-sonnet-4-6 | Balanced workhorse for production applications - speed, intelligence, and cost in equilibrium. | 1M / 128K | $3 / $15 |
| Claude Haiku 4.5claude-haiku-4-5 | Fastest and most economical - high-volume classification, extraction, routing, and quick chat. | 200K | $1 / $5 |
*Sonnet 5 introductory pricing runs through August 31, 2026; the standard $3 / $15 rate applies afterward. Anthropic's 1M-context models carry no long-context surcharge - a 900K-token request bills at the same per-token rate as a 9K one. Details: Fable 5 & Mythos 5 announcement.
Every Claude generation since 2023, what each one introduced, and how the naming works. Anthropic names its size tiers after forms of writing - the bigger the form, the bigger the model.
The smallest, fastest, cheapest tier - near-instant responses for chat, classification, extraction, and high-volume production work. Read more
The mid-size balanced tier and the default in the Claude apps - the best trade-off between intelligence, speed, and cost for most tasks. Read more
The flagship tier - maximum capability for the hardest coding, research, analysis, and long-running agent workflows. Read more
New in 2026: a tier above Opus, debuting with the Claude 5 family. Fable 5 is the generally available version; Mythos 5 is restricted to approved organizations.
Claude 1 launched in March 2023 alongside the lightweight Claude Instant, establishing Anthropic's "helpful, honest, harmless" approach. Claude 2 (July 2023) opened public access at claude.ai and expanded context to 100K tokens; Claude 2.1 (November 2023) doubled it to 200K and cut hallucination rates roughly in half. All are now retired.
The first generation released as a three-model family - Haiku, Sonnet, and Opus - introducing the size tiers still used today, plus vision (image understanding) across the lineup. Claude 3 Opus was briefly the strongest model in the industry.
Claude 3.5 Sonnet (June 2024, upgraded that October) became a developer favorite and introduced Artifacts and early computer use. Claude 3.7 Sonnet (February 2025) was the first hybrid reasoning model - one model that could answer instantly or "think" step-by-step on demand - and shipped alongside the first release of Claude Code.
Claude Opus 4 and Sonnet 4 (May 2025) were built for long-running, tool-using agent work, followed by rapid point releases: Opus 4.1 (August), Sonnet 4.5 (September - at launch, the best coding model in the world), Haiku 4.5 (October - near-frontier quality at a third of Sonnet's price), and Opus 4.5 (November). Context windows grew to 1M tokens on Sonnet during this era.
2026 brought fast iteration on the flagship line: Sonnet 4.6 became the production workhorse, while Opus advanced through 4.6, 4.7 (April - 1M context, high-resolution vision, self-verification), and 4.8 (May - adaptive thinking, effort controls, and an optional Fast Mode at 2× price for up to 2.5× speed).
The newest generation. Sonnet 5 (June 30, 2026) is the most agentic Sonnet yet and the new default on Free and Pro, at introductory pricing through August. Claude Fable 5 and Claude Mythos 5 - the first Claude 5 frontier models - share one underlying model in the new Mythos-class tier above Opus: Fable 5 is the most intelligent generally available Claude and adds safety measures for dual-use capabilities, while Mythos 5 omits those measures and is available only to approved organizations.
How to read version numbers: the family number (3, 4, 5) marks a new generation; point releases (4.5 → 4.6 → 4.8) are upgrades within it. Older models are eventually deprecated then retired from the API with advance notice - the current list of active model strings lives at docs.claude.com. Highlighted chips above mark models you can use today.
Subscriptions cover the Claude apps (web, mobile, desktop) and Claude Code. They are separate from the pay-per-token API. Usage works as a rolling ~5-hour budget, plus weekly caps on paid tiers.
Prices in USD before tax, subject to change - confirm at claude.com/pricing. When a paid plan's limit is reached, wait for the reset or buy extra usage at API token rates with a spending cap you control.
The Claude API (Claude Platform) is pay-as-you-go: no subscription, no monthly minimum. You buy prepaid credits - with optional auto-reload - and usage depletes them per token.
A token is roughly ¾ of an English word. Input (prompt + context) and output (Claude's reply) are billed separately per million tokens (MTok). Example: a Sonnet 4.6 call with 10K input and 2K output tokens costs about $0.06.
Batch API: 50% off everything for asynchronous jobs. Prompt caching: cache reads cost 10% of the input price (writes 1.25× for 5-minute TTL, 2× for 1-hour). Stacked, they can cut effective spend by 90%+.
Messages API, streaming, tool use / function calling, web search, code execution, MCP connectors, and the Agent SDK. US-only inference adds a 1.1× multiplier; Opus 4.8 Fast Mode runs at 2× standard rates.
Start at platform.claude.com · docs at docs.claude.com. Note: a Pro/Max subscription does not include API credits, and API credits don't unlock the chat apps - they are separate purchases.
One model family, many surfaces - everything below runs on the same Claude models.
The core assistant at claude.ai plus iOS, Android, and desktop apps - web search, deep research, file analysis, artifacts, projects, memory, and custom styles. Read more
An agentic coding tool for the command line, desktop app, VS Code, and JetBrains - delegate whole tasks, from bug fixes to features. Included with Pro, Max, and Team Premium seats, or run on an API key.
An agentic desktop app for non-developers: hand off research, analysis, and document work. Drives the other Claude tools and is reachable remotely from the mobile app.
A browsing agent that navigates sites, clicks, and fills out forms on your behalf.
A spreadsheet agent for formulas, cleanup, and models - and a slides agent that turns ideas into polished decks.
The Claude Platform for building apps and agents, with tool use, MCP, batching, and caching. Also available via Amazon Bedrock, Google Cloud, and Microsoft Foundry.
Claude runs everywhere - nothing to install for the web chat, and free official apps for every major platform. All of them share one account and one conversation history.
The fastest way in: open claude.ai/new in any browser, sign in, and chat. Full features - web search, artifacts, projects, uploads - with zero download.
The desktop app adds Cowork, Claude Code, quick-launch shortcuts, and the ability to work alongside local files and apps. Get it at claude.ai/download.
Free official apps on the App Store and Google Play - voice input, camera and photo analysis, and remote access to Claude Code and Cowork sessions. The publisher is Anthropic; skip lookalikes.
A Chrome extension that lets Claude browse, click, and fill out forms for you - agent mode for the open web.
Put Claude directly inside spreadsheets and slide decks for formulas, data cleanup, financial models, and deck building.
For developers: npm install -g @anthropic-ai/claude-code, then run claude in a project. Also available inside the desktop app, VS Code, and JetBrains IDEs.
Security tip: download only from claude.ai, the Apple App Store, or Google Play - never from third-party APK or installer sites.
How Claude stacks up against other frontier assistants in mid-2026. All four offer free tiers and a ~$20/month pro plan - the real differences are strengths, ecosystems, and API economics.
| Claude · Anthropic | ChatGPT · OpenAI | Gemini · Google | Grok · xAI | |
|---|---|---|---|---|
| Flagship models | Fable 5, Opus 4.8, Sonnet 5 | GPT-5.5 / GPT-5.4 | Gemini 3.1 Pro | Grok 4 |
| Consumer price | Free · $20 Pro · $100–200 Max | Free · $20 Plus · $200 Pro | Free · $20 Advanced | Free · ~$30 Premium+ |
| Flagship API · per 1M tokens | $5 / $25 (Opus 4.8) | $5 / $30 (GPT-5.5) | $2 / $12 (3.1 Pro)* | varies |
| Context window | Up to 1M tokens - no long-context surcharge | Up to 1M tokens | Up to 1M tokens - input price doubles above 200K | Large; varies by tier |
| Known strengths | Writing quality, coding & large codebases, instruction-following, agents (Code, Cowork), low hallucination rates, ad-free | Broadest ecosystem & integrations, best voice mode, image generation, versatile all-rounder | Multimodal - video & audio analysis, Google Workspace integration, low API cost | Speed, real-time knowledge from X, direct tone |
| Trade-offs noted by reviewers | Stricter usage limits than rivals; no native image generation | Can drift on long multi-part instructions; formulaic prose | Less consistent on complex code and nuanced writing | Smaller tool and app ecosystem |
*Gemini 3.1 Pro input rises to $4 / $18 above 200K-token prompts. Snapshot of independent reviews and vendor price pages, July 2026 - capabilities shift monthly, so test on your own tasks.
You care most about writing quality, serious coding on large codebases, precise instruction-following, long documents, or agentic workflows across terminal, browser, and Office. In a 134-person blind test published in 2026, Claude won 4 of 8 head-to-head rounds - the most of any model.
You want the most versatile all-rounder with the biggest third-party ecosystem, the best voice mode, and built-in image generation.
You live in Google Workspace, need video/audio understanding, or want the lowest frontier-class API rates. Many power users run two - Claude for depth plus one of the others for breadth.
From zero to first chat in about a minute. No credit card required for the Free plan.
Open claude.ai/new in any modern browser, or install the Claude app from the App Store or Google Play.
Continue with Google, or enter your email to receive a one-time login code - there's no password to remember. Verify, add your name, and you're in.
Type anything into the box. Attach PDFs, images, spreadsheets, or code with the paperclip; pick a model from the selector; toggle web search or research when you need current information.
Ask for an artifact ("build me an interactive budget calculator") to get documents, code, and mini-apps in a side panel. Group related chats and files into Projects. Set tone and format defaults in Settings → Preferences.
Seeing rate-limit messages more than once a week? Pro ($20/mo) is the usual next step - Settings → Billing, or claude.com/pricing. Developers can create an API key at platform.claude.com instead.
On Pro or above, install with npm install -g @anthropic-ai/claude-code, then run claude inside a project folder and log in with your Claude account.
Yes. The Free plan costs $0, needs no credit card, and includes the Sonnet 5 model, web search, file analysis, and artifacts on web, mobile, and desktop. Usage limits reset on a rolling ~5-hour window; paid plans raise those limits and add features.
Subscriptions (Free/Pro/Max/Team/Enterprise) are flat monthly fees for using the Claude apps and Claude Code. The API is pay-per-token access for developers building their own software. They're billed completely separately - a Pro subscription doesn't include API credits, and API credits don't upgrade your chat plan.
For everyday chat, the default Sonnet is excellent. Choose Opus 4.8 for the hardest coding, analysis, and agent tasks. Fable 5 sits above Opus for frontier-level reasoning (at 2× the price on the API). On the API, Haiku 4.5 is the go-to for high-volume, cost-sensitive workloads.
Tokens are the chunks models read and write - roughly ¾ of an English word each, so 1,000 tokens ≈ 750 words. The context window is how much a conversation can hold at once: up to 1 million tokens (several long novels' worth) on Fable 5, Opus 4.8, Sonnet 5, and Sonnet 4.6, and 200K on Haiku 4.5.
They're the first models in Anthropic's Claude 5 family and share the same underlying model in a new "Mythos-class" tier above Opus. Fable 5 is generally available and includes additional safety measures for dual-use capabilities; Mythos 5 is the same model without those measures, offered only to approved organizations. See Anthropic's announcement.
Anthropic describes limits as a rolling budget rather than a fixed message count - you get a warning as you approach the cap, then a reset time (sessions reset within about five hours, with weekly caps on paid tiers). Longer messages, bigger files, and pricier models consume the budget faster. Paid users can also buy extra usage at API rates when they hit a wall.
Claude products are ad-free - Anthropic doesn't display ads or let advertisers pay for placement in conversations (see Anthropic's policy). For current data and privacy settings, check Settings → Privacy in your account and support.claude.com.
There's nothing to forget: Claude uses passwordless login. Enter your email at claude.ai and a fresh one-time code is sent to you, or use "Continue with Google" if you signed up that way. If codes aren't arriving, check spam and confirm you're using the exact email you registered with; more help at support.claude.com.
Projects group chats and files around shared context. Artifacts are standalone outputs - documents, code, interactive apps - that open in a side panel and can be edited iteratively. Skills are reusable instruction folders that teach Claude best practices for specific tasks.
MCP (Model Context Protocol) is the open standard for connecting Claude to outside tools and data - Google Drive, Slack, Asana, GitHub, and hundreds more - via connectors you enable in settings.
Plans and prices: claude.com/pricing · Developer docs: docs.claude.com · Help center: support.claude.com · Product news: anthropic.com/news.
Start free at claude.ai - no credit card, no ads. Upgrade only when you feel a limit.