Back to products
Sonnet 4.6

Sonnet 4.6

The most capable Sonnet model yet

Website anthropic.com
Overview

What it is

We're an AI research company that builds reliable, interpretable, and steerable AI systems. Our first product is Claude, an AI assistant for tasks at any scale.

Intent

I need it when

Build and deploy AI agents at scale with balanced intelligence and cost

Sonnet 4.6 provides optimal balance of intelligence, cost, and speed. It supports managed agents at standard token rates ($0.08 per session-hour for active runtime) and integrates with Claude Code for building software, making it ideal for production agent deployments where cost-per-token matters.

Integrate AI into business workflows with predictable monthly costs

Sonnet 4.6 is available through Pro ($17-20/month), Team ($20-100/seat/month), and Enterprise plans with predictable billing. Users get access to Claude Code, Claude Cowork for delegating tasks, and connectors to Slack, Google Workspace, and Microsoft 365 for seamless workflow integration.

Process large volumes of asynchronous work with cost savings

Sonnet 4.6 supports batch processing at 50% discount for asynchronous workloads, reducing effective token costs to $1.50 input/$7.50 output per MTok. Ideal for scheduled tasks, research reports, and non-time-sensitive analysis that can be queued together.

Analyze data, write code, and solve complex problems without excessive API costs

Sonnet 4.6 offers mid-tier pricing ($3 input/$15 output per MTok) positioned as the optimal balance between capability and expense. It supports code execution, file creation, and data visualization through artifacts, enabling developers and analysts to tackle substantive work efficiently.

Drop

Not a fit when

  • User requires the absolute highest intelligence for complex reasoning tasks; Opus 4.8 is positioned as the most capable model for agents and coding
  • User needs the fastest inference speed with minimal latency; Haiku 4.5 is optimized for speed and cost-efficiency
  • User operates under strict budget constraints with minimal token usage; Haiku 4.5 offers lower pricing at $1/MTok input and $5/MTok output
  • User requires US-only data residency; pricing increases 1.1x for US-only inference which may exceed budget
  • User needs real-time processing with guaranteed sub-second response times; batch processing tier is for asynchronous workloads only
Commercials

Pricing

API usage-based pricing per million tokens (MTok). Input $3/MTok, Output $15/MTok. Also available through subscription plans (Free, Pro $17-20/month, Max from $100/month, Team $20-100/seat/month, Enterprise custom). View pricing