Skip to content

Anthropic Claude Models — Complete Guide

🟣 Anthropic Claude Models — Complete Guide

Section titled “🟣 Anthropic Claude Models — Complete Guide”

Anthropic organises its Claude models into three named tiers that have remained consistent across every generation: Opus (maximum capability), Sonnet (balanced), and Haiku (fast and cost-efficient). Each generation raises the bar across all three tiers.


📅 Generation🔮 Opus⚖️ Sonnet⚡ Haiku🗓️ Year
Claude 3Claude 3 OpusClaude 3 SonnetClaude 3 HaikuMar 2024
Claude 3.5Claude 3.5 Sonnet (Jun + Oct 2024)Claude 3.5 HaikuOct 2024
Claude 3.7Claude 3.7 SonnetFeb 2025
Claude 4Claude Opus 4Claude Sonnet 4May 2025
Claude 4.5Claude Opus 4.5Claude Sonnet 4.5Claude Haiku 4.5Sep–Nov 2025
Claude 4.6Claude Opus 4.6Claude Sonnet 4.6Feb 2026
Claude 4.7Claude Opus 4.7Apr 2026
Claude 4.8Claude Opus 4.8 ✅ CurrentMay 2026

These are the models in active use and recommended for new projects:

🤖 Model🏷️ Tier📅 Released📄 Context🎯 Best For
🔮 Claude Opus 4.8FlagshipMay 20261M tokensComplex reasoning, long-horizon agents
⚖️ Claude Sonnet 4.6BalancedFeb 2026200K tokensDay-to-day coding, writing, production
⚡ Claude Haiku 4.5FastOct 2025200K tokensClassification, routing, high-volume tasks

The Opus tier is Anthropic’s flagship — the strongest reasoning, the deepest coding ability, and the longest context handling.

🔮 Claude Opus 4.8 — Current Flagship (May 2026)
🔧 Property📋 Detail
📅 ReleasedMay 28, 2026
📄 Context window1,000,000 tokens
🎯 FocusComplex reasoning · long-horizon agentic coding · professional work
🌐 AccessAnthropic API · Claude.ai
🏆 BenchmarksState-of-the-art on SWE-bench coding and agentic evaluations

✅ Ideal for:

  • 🚨 Production incident debugging where failure is expensive
  • 🏗️ Architecture decisions and migration planning
  • 📁 Large cross-file refactors requiring coherent multi-step plans
  • 👀 Reviewing high-stakes pull requests for security and correctness
  • 🤖 Long-horizon agentic tasks that span many tool calls and steps
  • 📄 Analysing very long documents (contracts, research papers, codebases)

💡 Example:

“Review this entire microservices codebase (attached). Identify the top three architectural risks, propose a phased remediation plan, and highlight which services must be touched first to unblock the others.”

📅 Earlier Opus Models
🤖 Model📅 Released📋 Notes
Claude Opus 4.7Apr 2026Incremental improvement over 4.6
Claude Opus 4.6Feb 2026Preferred by independent raters over comparable models
Claude Opus 4.5Nov 2025First Opus with significantly expanded agentic ability
Claude Opus 4May 2025First of the Claude 4 generation
Claude 3 OpusMar 2024Original flagship — strong reasoning, 200K context

Earlier Opus versions are still accessible via the API but newer generations are recommended for new projects.


The Sonnet tier is the recommended default for most production use cases — strong enough for hard tasks, fast and cost-efficient enough for everyday work.

⚖️ Claude Sonnet 4.6 — Current Balanced Model (Feb 2026)
🔧 Property📋 Detail
📅 ReleasedFebruary 2026
📄 Context window200,000 tokens
🎯 FocusCoding · writing · analysis · production assistant workflows
🌐 AccessAnthropic API · Claude.ai · Amazon Bedrock · Google Cloud Vertex AI

✅ Ideal for:

  • 💻 Day-to-day coding: adding features, fixing bugs, writing tests
  • 📝 Writing and editing: technical docs, summaries, emails
  • 🔍 Code review and explanation
  • 🤖 Building production AI assistants and agents
  • 📊 Data analysis and structured extraction

💡 Example:

“Add input validation and rate limiting to this Express.js API endpoint. Write Jest tests for the happy path and three error cases. Update the OpenAPI spec.”

📅 Earlier Sonnet Models
🤖 Model📅 Released📋 Notable Feature
Claude Sonnet 4.5Sep 2025Improved agentic performance
Claude Sonnet 4May 2025First Claude 4 generation Sonnet
Claude 3.7 SonnetFeb 2025Introduced extended thinking / hybrid reasoning
Claude 3.5 Sonnet v2Oct 2024Major upgrade — computer use capability
Claude 3.5 SonnetJun 2024Landmark release — outperformed Opus 3 at Sonnet price
Claude 3 SonnetMar 2024First Sonnet, balanced speed and quality

📌 Notable milestone: Claude 3.5 Sonnet (June 2024) was the first Anthropic model to outperform the previous generation’s Opus tier — a landmark in the efficiency progression.


⚡ Haiku Tier — Fast and Cost-Efficient

Section titled “⚡ Haiku Tier — Fast and Cost-Efficient”

The Haiku tier is designed for tasks where speed and cost matter more than maximum reasoning depth — classification, routing, extraction, and simple transformations at scale.

⚡ Claude Haiku 4.5 — Current Fast Model (Oct 2025)
🔧 Property📋 Detail
📅 ReleasedOctober 2025
📄 Context window200,000 tokens
🎯 FocusSpeed · cost efficiency · high-volume tasks
🌐 AccessAnthropic API · Claude.ai · Amazon Bedrock

✅ Ideal for:

  • 🏷️ Classifying support tickets, emails, or reviews at scale
  • 🔀 Routing user requests to the right handler
  • 🔎 Extracting structured fields from short text
  • ✏️ Rewriting short content (subject lines, summaries)
  • 📱 Latency-sensitive applications

💡 Example:

“Classify each of these 10,000 support tickets as: billing, login, technical, or feedback. Return a JSON array with ticket ID and label.”

📅 Earlier Haiku Models
🤖 Model📅 Released📋 Notes
Claude 3.5 HaikuNov 2024First 3.5 Haiku — faster than 3 Haiku with better quality
Claude 3 HaikuMar 2024Original Haiku — lightest and fastest Claude 3
Claude InstantMar 2023Pre-Claude 3 era fast model

🧠 Key Claude Capabilities (All Current Models)

Section titled “🧠 Key Claude Capabilities (All Current Models)”
✅ Capability📋 Detail
👁️ VisionUnderstands images, screenshots, charts, and diagrams
📄 Long contextOpus 4.8: 1M tokens · Sonnet 4.6 + Haiku 4.5: 200K tokens
🤖 Tool useCan call functions and external APIs in agentic workflows
💻 CodeStrong across Python, TypeScript, JavaScript, Go, Java, C#, Rust
🌍 MultilingualUnderstands and generates in major world languages
🛡️ SafetyBuilt-in Constitutional AI — designed to be helpful, harmless, and honest

🎯 Task✅ Best Model💡 Why
🏗️ Architecture design or migration planningOpus 4.8Deep reasoning + very long context
🚨 Production incident debuggingOpus 4.8Maximum reasoning on high-stakes problems
📁 Large cross-file refactorOpus 4.8Needs coherent multi-step planning
👀 Risky pull request reviewOpus 4.8Thorough bug detection across all files
💻 Everyday coding and feature workSonnet 4.6Strong quality at reasonable cost
📝 Writing, summarisation, documentationSonnet 4.6Excellent balance of quality and speed
🤖 Production AI assistant (chat)Sonnet 4.6Best default for most apps
🏷️ Classification or extraction at scaleHaiku 4.5Fastest and cheapest for structured tasks
🔀 Request routingHaiku 4.5Simple judgment at low cost per call

🏷️ Platform🤖 Models Available
🟣 Anthropic API (api.anthropic.com)All current Claude models
💬 Claude.aiOpus, Sonnet, Haiku (product tiers)
🟠 Amazon BedrockClaude Opus, Sonnet, Haiku
🔵 Google Cloud Vertex AIClaude Opus, Sonnet, Haiku
🔀 OpenRouterClaude models via unified API