Access · Pricing · API · Platforms

Get Started

Sonnet 4.6 is available now across every Claude plan, every major cloud platform, and via API.

Already on claude.ai Free or Pro? You're already on Sonnet 4.6. It became the default February 17, 2026. No action needed.

Access Options

Where To Find It

💬

Claude.ai

Web, mobile, desktop chat. Default for Free and Pro plans since Feb 17, 2026.

FreeProTeamEnterprise

Open Claude.ai ↗

⌨️

Claude Code

CLI agentic coding tool. Where the 70% developer preference test was run.

# Install
npm install -g @anthropic-ai/claude-code
# Launch
claude
      

Docs ↗

🏢

Claude Cowork

Desktop productivity tool. File and task automation. Default: Sonnet 4.6.

FreePro

Claude Cowork ↗

🔌

Claude API

Direct API access. Model ID: claude-sonnet-4-6. Same pricing as Sonnet 4.5.

API Docs ↗

☁️

Amazon Bedrock

Available now on AWS via Amazon Bedrock for teams in the AWS ecosystem.

AWS Bedrock ↗

🔷

Google Vertex AI

Available via Google Cloud Vertex AI for GCP-ecosystem teams.

Vertex AI ↗

Pricing

Same Price. Better Model.

Sonnet 4.6 costs exactly the same as Sonnet 4.5. Every developer building with Sonnet gets a free intelligence upgrade.

Claude 4.6 Family — API Pricing

Per million tokens, input and output

Source: claude.com/pricing#api. Haiku pricing approximate.

Sonnet 4.6 API Pricing

per million input tokens

$15

per million output tokens

Unchanged from Sonnet 4.5 — zero migration cost

vs. Opus 4.6 — 40% Cheaper

Sonnet 4.6 is 40% cheaper per token than Opus 4.6, while matching Opus 4.6 on OfficeQA and beating the previous Opus 4.5 flagship in 59% of developer coding tests.

Opus 4.6 Input

$5 / M

Opus 4.6 Output

$25 / M

💰

What "Same Price" Means

If you were spending $1,000/month on Sonnet 4.5, you're now spending $1,000/month on a model that beats the previous flagship in developer preference tests. Value per dollar increased automatically on February 17, 2026.

Developer Quick Start

API Code Examples

Minimal Python

import anthropic

client = anthropic.Anthropic()
message = client.messages.create(
  model="claude-sonnet-4-6",
  max_tokens=1024,
  messages=[{
    "role": "user",
    "content": "Hello!"
  }]
)
print(message.content)
      

With Adaptive Thinking

# Claude decides when to reason deeply
message = client.messages.create(
  model="claude-sonnet-4-6",
  max_tokens=16000,
  thinking={
    "type": "enabled",
    "budget_tokens": 10000
  },
  messages=[{
    "role": "user",
    "content": "Analyze this codebase..."
  }]
)
      

With Web Search (Now GA)

# Auto-filters search results
message = client.messages.create(
  model="claude-sonnet-4-6",
  max_tokens=2048,
  tools=[{
    "type": "web_search_20250305",
    "name": "web_search"
  }],
  messages=[{
    "role": "user",
    "content": "Latest AI news today?"
  }]
)
      

Model Identifiers

# Pinned — recommended for production:
model = "claude-sonnet-4-6"

# Auto-update to latest Sonnet:
model = "claude-sonnet-4-6-latest"
      

📌

Pin to claude-sonnet-4-6 for reproducible behavior in production. Use -latest only if you want future Sonnet updates automatically applied.

Platform Updates · February 17, 2026

What Else Shipped

🧠 Adaptive Thinking

Claude decides when deeper reasoning is warranted. At default (high) effort, extended thinking activates automatically for complex steps — no binary ON/OFF required.

Docs ↗

📦 Context Compaction (Beta)

Automatically summarizes older context approaching limits, enabling continuously running agents. Configurable threshold. Available in API and claude.ai.

Docs ↗

🔍 Improved Web Search

Web search and fetch now auto-write and execute code to filter results — keeping only relevant content in context. Better quality, better token efficiency.

Docs ↗

📊 Claude in Excel — MCP

MCP connectors now supported: S&P Global, LSEG, Daloopa, PitchBook, Moody's, FactSet. Pull external data into Excel without leaving the spreadsheet. Pro, Max, Team, Enterprise.

Docs ↗

🔧 Newly GA API Tools

These moved from beta to generally available alongside Sonnet 4.6:

Code Execution Memory Tool Programmatic Tool Calling Tool Search Tool Use Examples

🎁 Free Tier Upgrades

Features previously Pro-only, now free:

File Creation Connectors Skills Context Compaction

Official Resources

Every Link You Need

Resource	Description	Link
Official Announcement	Full launch post from Anthropic	anthropic.com ↗
System Card	Full benchmark data + safety evaluation	system-card ↗
Pricing Page	All plan and API pricing	claude.com/pricing ↗
API Models Overview	Model identifiers and capabilities	platform docs ↗
Adaptive Thinking	How to use the new thinking system	docs ↗
Context Compaction	How to configure compaction	docs ↗
Prompt Injection Guide	Safety guide for computer use deployments	security ↗
OSWorld Benchmark	Computer use benchmark methodology	os-world.github.io ↗
Vending-Bench Arena	Business simulation benchmark	andonlabs.com ↗
OfficeQA / GDPVal	Enterprise document benchmark	artificialanalysis.ai ↗
Claude Code	CLI coding agent product page	product page ↗
Claude Cowork	Desktop productivity tool	product page ↗

Ready?

Opus intelligence.
Sonnet price.

Claude Sonnet 4.6 is live now. If you're already on claude.ai, you're already using it.

Try It Now ↗ API Docs ↗ See All Benchmarks

Get Started

Where To Find It

Claude.ai

Claude Code

Claude Cowork

Claude API

Amazon Bedrock

Google Vertex AI

Same Price. Better Model.

Sonnet 4.6 API Pricing

vs. Opus 4.6 — 40% Cheaper

API Code Examples

Minimal Python

With Adaptive Thinking

With Web Search (Now GA)

Model Identifiers

What Else Shipped

🧠 Adaptive Thinking

📦 Context Compaction (Beta)

🔍 Improved Web Search

📊 Claude in Excel — MCP

🔧 Newly GA API Tools

🎁 Free Tier Upgrades

Every Link You Need

Opus intelligence.Sonnet price.

Opus intelligence.
Sonnet price.