Claude Sonnet 4.5 token cost & pricing

Claude Sonnet 4.5 is Anthropic's balanced frontier model and the strongest widely-available option for agentic coding and long-horizon tool use. A 200K context window plus a 64K max output makes it practical for tasks that produce large, coherent artifacts such as full file rewrites or multi-step reports.

Claude Sonnet 4.5 specifications

What Claude Sonnet 4.5 costs in practice

Four common workloads, priced at list rates. Monthly figures assume 30 days.

How many requests does $100 buy?

Assuming a 1,000-token prompt and a 500-token response per request.

How Claude Sonnet 4.5 pricing actually works

At $3 per million input and $15 per million output, Sonnet sits above GPT-5 on both sides — about 2.4x on input and 1.5x on output. Cached input at $0.30 per million is a 90% discount, which is the single most important lever for Anthropic models: agentic workloads resend large system prompts and tool definitions constantly.

When to use Claude Sonnet 4.5

Pick Sonnet for coding agents, refactoring tasks, long document analysis and anything where instruction-following consistency matters more than price. For high-volume classification or short chat, Haiku 4.5 costs about a third as much.

How to cut your Claude Sonnet 4.5 bill

Enable prompt caching first — it is the difference between a viable and an unviable agent budget. Then pin tool definitions to the cached prefix and keep the volatile part of the prompt at the end, so cache hits actually land.

Other Anthropic models

Cheaper alternatives to Claude Sonnet 4.5

• GPT-4oOpenAI31% cheaper
• GPT-5OpenAI38% cheaper
• Gemini 2.5 ProGoogle38% cheaper
• Mistral Large 2Mistral56% cheaper