Claude
The AI built to be safe, honest, and helpful.
Four tiers, frontier context
Haiku, Sonnet, and Opus share a 200K token context window. Claude Fable 5 — Anthropic's new flagship — extends that to 1M tokens. Choose your tier by the cost-performance trade-off your application demands.
Claude Haiku 4.5
Fastest Claude for high-volume tasks
Context
200K
Speed
⚡ fastest
Input
$0.80 / 1M tokens
Output
$4 / 1M tokens
Best for
Key capabilities
- Lowest latency in the Claude family
- Cost-optimized for production scale
- 200K token context window
- Strong instruction following
- Ideal for agentic sub-tasks that require speed
Claude Sonnet 4.6
The sweet spot of intelligence and speed
Context
200K
Speed
◐ balanced
Input
$3 / 1M tokens
Output
$15 / 1M tokens
Best for
Key capabilities
- Top performance on SWE-bench coding evals
- 200K token context for multi-document tasks
- Accurate multi-step agentic execution
- Nuanced instruction following and tone control
- Best cost-to-capability ratio for production apps
Claude Opus 4.7
Maximum intelligence with extended thinking
Context
200K
Speed
🧠 most capable
Input
$15 / 1M tokens
Output
$75 / 1M tokens
Best for
Key capabilities
- Extended Thinking mode for deep chain-of-thought
- Highest accuracy on complex reasoning benchmarks
- Ideal for long-horizon autonomous agent tasks
- Superior performance on graduate-level STEM problems
- Best-in-class for nuanced multi-step decision making
Claude Fable 5
Anthropic's most capable model — frontier agentic reasoning
Context
1M
Speed
🧠 most capable
Input
$10 / 1M tokens
Output
$50 / 1M tokens
Best for
Key capabilities
- 95% on SWE-bench Verified — state of the art at launch
- 1M-token context window with up to 128K output tokens
- Always-on adaptive thinking — no separate 'extended thinking' toggle
- Sustains autonomous multi-day tasks with self-verification and sub-agent delegation
- Memory tool, code execution, and programmatic tool calling built in
* Prices per 1M tokens, approximate as of 2025. Verify current rates at anthropic.com/pricing.
Anthropic
Claude Code
An agentic coding CLI that lives in your terminal. Claude Code reads and edits files, runs shell commands, manages git, and executes multi-step software engineering tasks autonomously — without leaving your workflow.
VS Code & JetBrains integration
Embedded Claude Code panel inside your editor — run tasks, view diffs, and accept changes inline.
MCP server support
Connect any Model Context Protocol server to give Claude Code access to external APIs, databases, and tools.
Multi-agent workflows
Orchestrate parallel sub-agents for large-scale tasks: one agent per PR, feature, or test suite.
Install
$ npm install -g @anthropic-ai/claude-code
$ claude
API · TypeScript SDK
import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic();
const msg = await client.messages.create({
model: 'claude-sonnet-4-6',
max_tokens: 1024,
messages: [{
role: 'user',
content: 'Hello, Claude!',
}],
});Every way to access Claude
Claude.ai
The flagship Claude chat interface
A polished web and mobile chat interface giving direct access to the Claude model family. Supports Projects for persistent context, artifact rendering, file uploads, vision, and voice.
- Projects with persistent memory across conversations
- Artifact preview for code, documents, and SVGs
- File and image upload with vision understanding
- Free tier available with Claude access
Claude API
Full programmatic access via Anthropic's API
Enterprise-grade REST API giving developers access to all Claude models with tool use, streaming, batch processing, prompt caching, and the full suite of capabilities for production applications.
- Access all Claude models with a single API key
- Prompt caching for up to 90% cost reduction on repeated context
- Streaming, tool use, and structured output support
- Python and TypeScript SDKs with full type safety
Claude Code
Agentic AI coding assistant for the terminal
A CLI tool that brings Claude directly into your development workflow. Claude Code can read and edit files, run terminal commands, manage git, and complete multi-step software engineering tasks autonomously.
- Reads, writes, and refactors entire codebases
- Runs shell commands, tests, and git operations
- MCP server support for custom tool integrations
- Remote Control for browser-based agent sessions
Claude iOS App
Claude on the go — with Remote Control support
The native iOS app for Claude gives you access to conversations on iPhone and iPad, with voice input, image understanding, and Remote Control capability to manage Claude Code sessions from your phone.
- Full Claude model access on iPhone and iPad
- Voice input for hands-free conversation
- Remote Control: monitor and guide Claude Code sessions
- Projects sync with claude.ai for continuity
What Claude can do
Core technical capabilities available across the API — from vision and tool use to extended thinking and the Model Context Protocol.
Extended Thinking
Docs
Claude Opus 4.7 can reason through problems step-by-step before answering, spending more compute tokens on hard problems to produce more accurate, transparent conclusions.
Tool Use
Docs
Define custom tools (functions) that Claude can invoke mid-conversation. Enables reliable structured data extraction, API calls, and dynamic agentic behavior.
Vision
Docs
All Claude models can analyze images — screenshots, diagrams, documents, photographs. Upload images via base64 or URL for multi-modal reasoning.
Computer Use
Docs
Claude can interact with a computer: move the mouse, click, type, take screenshots, and navigate GUIs autonomously. Enables fully automated desktop and browser workflows.
Multi-turn Conversations
The Messages API supports full conversation history, enabling coherent multi-turn dialogues, project-scoped memory, and complex back-and-forth reasoning sessions.
Batch API
Docs
Process thousands of requests asynchronously at up to 50% lower cost. Ideal for offline evaluation, data labeling, content generation, and large-scale analysis pipelines.
Model Context Protocol
Docs
An open standard for connecting AI models to external data sources and tools. MCP servers expose resources, prompts, and tools that Claude can discover and use at runtime.
Prompt Caching
Docs
Cache frequently reused context (system prompts, documents, tool definitions) between API calls. Reduces latency by up to 85% and costs by up to 90% for cache hits.
Resources & references