Claude Model Family
Claude Haiku
Fast · Efficient
The fastest and most compact Claude model. Ideal for high-throughput tasks, real-time interactions, and cost-sensitive workloads where speed matters most.
Context window200K tokens
Latestclaude-haiku-4-5
Best forSummarization, classification, quick Q&A
Claude Sonnet
Balanced · Capable
The ideal balance of intelligence and speed. Sonnet excels at complex reasoning, coding, analysis, and writing — the go-to model for most production use cases.
Context window200K tokens
Latestclaude-sonnet-4-6
Best forCoding, analysis, complex tasks
Claude Opus
Powerful · Deep Reasoning
Anthropic's most intelligent model. Opus handles the most demanding tasks — research synthesis, expert-level coding, nuanced writing, and multi-step reasoning chains.
Context window200K tokens
Latestclaude-opus-4-6
Best forResearch, hard problems, expert tasks
Core Capabilities
What Claude Can Do
- 💻Coding — writes, reviews, debugs, and explains code across 50+ languages
- 📝Writing — drafts, edits, and refines prose from emails to long-form essays
- 🔍Analysis — breaks down complex documents, data, and arguments
- 🧮Math & reasoning — solves problems step-by-step with chain-of-thought
- 🌐Multilingual — reads and writes in dozens of languages
- 🖼️Vision — understands images, diagrams, charts, and screenshots
- 🛠️Tool use — calls functions, searches the web, writes & runs code in sandboxes
- 📂Long documents — reads entire codebases or books in a single context window
Anthropic Timeline
2021
Anthropic founded — Dario Amodei, Daniela Amodei, and seven others leave OpenAI to focus on AI safety
2022
Constitutional AI paper — Anthropic publishes the CAI method for training helpful, harmless, honest models
Mar 2023
Claude 1 launches — first public Claude model via API and Claude.ai
Mar 2024
Claude 3 family — Haiku, Sonnet, Opus debut; Opus tops MMLU and HumanEval benchmarks
Jun 2024
Claude 3.5 Sonnet — sets new state-of-the-art on coding; artifacts and computer use previewed
2025
Claude 4 generation — Sonnet 4.6 and Opus 4.6 with extended thinking, Claude Code CLI, and agent capabilities
Constitutional AI & Safety
The Constitution
A set of principles Claude is trained to follow — be helpful, harmless, and honest. Claude critiques and revises its own outputs against these principles during training.
RLHF + RLAIF
Reinforcement learning from human feedback (RLHF) is augmented with AI feedback (RLAIF) — Claude evaluates its own responses to scale safety training without exhausting human annotators.
Harmlessness
Claude refuses requests that would cause real-world harm. It's trained to recognize nuanced harmful intent, not just keyword lists, and explain its reasoning when declining.
Interpretability
Anthropic's research team works on mechanistic interpretability — understanding exactly which model weights and circuits produce specific behaviors, making AI more auditable.
Honesty
Claude is trained to be calibrated — expressing appropriate uncertainty, not making up facts, and flagging when it doesn't know something rather than confabulating.
Red Teaming
Before each release, Anthropic and external red teamers attempt to elicit harmful outputs. Findings feed back into training to patch vulnerabilities before public release.
API Quick Start
# Install the SDK
pip install anthropic
# Basic message (Python)
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
messages=[
{"role": "user", "content": "Explain neural networks in 3 sentences."}
]
)
print(message.content[0].text)
# With system prompt
message = client.messages.create(
model="claude-opus-4-6",
max_tokens=2048,
system="You are an expert software architect.",
messages=[
{"role": "user", "content": "Design a microservices architecture for a food delivery app."}
]
)
Model Comparison
| Capability | Haiku 4.5 | Sonnet 4.6 | Opus 4.6 |
|---|---|---|---|
| Speed | ⚡ Fastest | ⚡ Fast | 🐢 Slower |
| Complex reasoning | ★★★ | ★★★★ | ★★★★★ |
| Coding (HumanEval) | ~75% | ~90% | ~93% |
| Context window | 200K | 200K | 200K |
| Vision / images | ✓ | ✓ | ✓ |
| Tool / function use | ✓ | ✓ | ✓ |
| Extended thinking | ✗ | ✓ | ✓ |
| Cost (relative) | Lowest | Medium | Highest |
| Best use case | Real-time, bulk | General production | Hard problems |
Claude Code
What is Claude Code?
- 🖥️An official CLI tool from Anthropic that runs Claude directly in your terminal
- 📁Reads your entire codebase, edits files, runs shell commands, and commits to git
- 🔁Agentic loop — Claude plans, acts, observes results, and iterates until the task is done
- 🪝Configurable hooks let you automate behaviors on tool use events
- 🧩MCP (Model Context Protocol) support for custom tool integrations
- ⌨️IDE extensions for VS Code and JetBrains bring Claude Code into your editor
Key Commands
- 💬
claude— start an interactive session - 📤
claude "fix the login bug"— one-shot task - 📋
claude --help— list all flags - 🔑
ANTHROPIC_API_KEY— set your API key - 📂
CLAUDE.md— project-level instructions file - ⚡
/fast— toggle fast mode (Opus 4.6) - 🗑️
/clear— clear conversation context