Claude 5 Sonnet

Claude 5 Sonnet

**Claude Sonnet 5** is the next generation of Anthropic's Sonnet model family, delivering the best combination of speed and intelligence at the Sonnet tier. It is a drop-in upgrade for Claude Sonnet 4.6 with significant capability gains — especially in coding and agentic tasks — at the same per-token price. ## Key Capabilities - **1M token context window by default**: Claude Sonnet 5 supports a 1 million token context window as both the default and maximum, enabling reasoning over massive codebases, document sets, and long-running agentic sessions. - **128K max output tokens**: Generate long-form responses, detailed plans, or extensive code in a single completion. - **Adaptive thinking by default**: The model dynamically decides when and how much to think based on task complexity — no manual budget tuning required. Adaptive thinking is on by default; pass `thinking: {type: "disabled"}` to turn it off. - **Agentic coding & tool use**: The largest gains over Claude Sonnet 4.6 are in coding and agentic tasks, making Sonnet 5 a strong fit for autonomous developer agents, long-horizon workflows, and tool-heavy applications. - **Cybersecurity safeguards**: The first Sonnet-tier model with real-time cybersecurity safeguards, refusing requests involving prohibited or high-risk cybersecurity topics. ## What's New - **New tokenizer**: Claude Sonnet 5 uses a new tokenizer that produces approximately 30% more tokens for the same input text compared to Claude Sonnet 4.6. Per-token pricing is unchanged, but token budgets and `max_tokens` limits should be revisited. - **Manual extended thinking removed**: The legacy `thinking: {type: "enabled", budget_tokens: N}` parameter is no longer supported — use adaptive thinking instead. - **Sampling parameters disabled**: Setting `temperature`, `top_p`, or `top_k` to non-default values returns a 400 error. Use system-prompt instructions to guide model behavior. ## What It's Best For Claude Sonnet 5 is purpose-built for production workloads that need more capability than Sonnet 4.6 provides without moving to an Opus-class model. Ideal for agentic coding, complex tool-using assistants, long-context retrieval, document analysis at scale, and customer-facing workflows where execution quality and response polish are critical. ## Pricing Standard pricing is $3 per million input tokens and $15 per million output tokens. Introductory pricing of $2/$10 per million input/output tokens is in effect through August 31, 2026.

Explore similar models

Anthropic

Claude 5 Fable

Anthropic's most intelligent model, setting new standards in agentic coding, complex reasoning, and professional knowledge work.

$10.00/MTok

$50.00/MTok

1M

Anthropic

Claude 4.8 Opus

Anthropic's most intelligent model, setting new standards in agentic coding, complex reasoning, and professional knowledge work.

$5.00/MTok

$25.00/MTok

1M

Anthropic

Claude 4.7 Opus

Anthropic's most intelligent model, setting new standards in agentic coding, complex reasoning, and professional knowledge work.

$5.00/MTok

$25.00/MTok

1M

Google

Gemini 3.1 Pro

Google's frontier reasoning model delivering enhanced software engineering, agentic reliability, and multimodal intelligence across a 1M-token context window.

$2.00/MTok

$12.00/MTok

1.0M

OpenAI

GPT‑5.2 Pro

The most capable model series yet for professional knowledge work.

$21.00/MTok

$168.00/MTok

400K

Qwen

Qwen3.6 Flash

Alibaba's fast, cost-efficient vision-language model with a 1M token context window, excelling at agentic coding, deep reasoning, and spatial visual understanding.

$0.25/MTok

$1.50/MTok

1M