Claude Sonnet 5 pricing is simpler than many small businesses expected. When Anthropic launched Claude Sonnet 5 in summer 2026, it charged an introductory $2 per million input tokens and $10 per million output tokens, and said the price would rise to $3/$15 on September 1, 2026. In August, Anthropic cancelled that increase, so the lower price became the standard rate.
If you rushed to lock in credits or rebuild workflows before September, you did not need to. This guide explains the current Claude Sonnet 5 pricing, what it costs in real use, and six proven ways to keep your bill low.
Table of Contents
Claude Sonnet 5 Pricing at a Glance
| Item | Price per 1M tokens |
|---|---|
| Input | $2 |
| Output | $10 |
| Cached input (reads) | Much lower than standard input; see Anthropic’s pricing page |
| Batch API | Discounted versus standard requests |
These are the reported list prices for the API. Anthropic adjusts its lineup often, and newer models such as Sonnet 5.5 may have their own pricing, so always confirm on the official Claude pricing page before you budget.
What Happened to the September Price Increase
The original plan was a 50% increase on both input and output tokens from September 1. Anthropic reversed that decision in August and updated its pricing documentation to say the scheduled increase would not happen. Every other line on the rate card stayed the same.
For small businesses, that means no deadline and no need to pre-buy credits. Plan your spending around the $2/$10 rate and review it when Anthropic releases new models.
How Claude Sonnet 5 Pricing Compares
Claude Sonnet 5 pricing sits below the flagship Opus models and above the small, fast Haiku models. Older Sonnet models remain available at their own rates, which were higher in some cases, so moving to Sonnet 5 can reduce costs as well as improve quality.
For a full comparison with OpenAI models, see our guide to Claude vs OpenAI for n8n automation.
What It Costs in Practice
Most small business tasks use far fewer tokens than people expect. A few rough examples at $2/$10:
| Task | Approximate tokens | Approximate cost |
|---|---|---|
| Reply to one customer email | 2,000 in / 300 out | Under $0.01 |
| Summarize a 20-page document | 15,000 in / 1,000 out | About $0.04 |
| 1,000 product descriptions | 200,000 in / 150,000 out | About $1.90 |
Costs climb when you send long documents on every request, keep huge chat histories, or let agents loop for many steps. That is where the savings below matter.
6 Proven Ways to Lower Your Bill
- Use prompt caching: if every request repeats the same long instructions or documents, caching cuts the cost of that repeated input sharply.
- Use the Batch API: for work that does not need an instant answer, such as overnight reports, batch requests are discounted.
- Route simple tasks to Haiku: classification, tagging and short replies rarely need Sonnet.
- Trim your inputs: send only the relevant part of a document, not the whole file.
- Limit output length: set a sensible maximum and ask for concise answers; output tokens cost five times more than input.
- Set spend limits and alerts: use the usage limits in the Claude Console so a broken workflow cannot run up a large bill.
Should You Switch Models?
If you already use Sonnet 4.x, Claude Sonnet 5 pricing makes switching worth testing: run your real prompts through both, compare quality and cost, and move over workflow by workflow. If you use Opus for everyday tasks, test whether Sonnet 5 is good enough; the savings can be large.
The Bottom Line
The feared September price hike never happened. Claude Sonnet 5 pricing stays at $2 per million input tokens and $10 per million output tokens, which makes it one of the best value models for business automation. Use caching, batching and model routing to cut costs further, and check Anthropic’s pricing page whenever a new model ships.
New to the API? Start with our Claude API tutorial.