Claude Sonnet 5

Claude Sonnet 5

Anthropic's mid-tier model in the Claude 5 family. Replaces claude-sonnet-4-6 as the default model in claude-code and API agentic workflows. Part of the Claude 5 generation, which introduces a restructured tier naming — Sonnet and claude-fable-5 — departing from the prior Haiku / Sonnet / Opus hierarchy.

Positioning

Sonnet 5 continues the Sonnet role: the primary workhorse for agentic coding, tool use, and autonomous sessions. Designed for professional API use where sustained capability across long contexts is the core requirement. claude-fable-5 handles the highest-complexity reasoning tasks; Sonnet 5 covers the vast majority of production use cases.

Context Window

1M tokens. Max output: 128K tokens.

Thinking

Supports extended-thinking with both adaptive mode (model decides) and manual budget_tokens configuration — consistent with claude-sonnet-4-6's approach, unlike Opus 4.8 which is adaptive-only.

Strengths

  • Agentic coding: Default in claude-code, positioned as the primary model for autonomous coding sessions and multi-agent-setup orchestration.
  • Extended thinking: Full manual and adaptive extended-thinking support with configurable reasoning budgets.
  • Cost-efficiency: Mid-tier pricing makes it viable for high-volume agentic work that would be prohibitive at Fable-tier cost.
  • Tool use: Reliable tool-use and mcp integration across structured and semi-structured tasks.

Use Cases

Best for: autonomous coding in claude-code, multi-step agentic-workflows, computer-use automation, API pipelines requiring sustained capability at reasonable cost.

Not ideal for: highest-complexity reasoning or analysis (use claude-fable-5), quick lightweight completions where latency is the only concern.

Availability

Claude API, AWS Bedrock, Google Vertex AI, Microsoft Foundry, claude.ai (paid plans), claude-code, GitHub Copilot.

Related

claude-fable-5 · claude-sonnet-4-6 · claude-opus-4-8 · extended-thinking · agentic-workflows · claude-code · computer-use

Sources