Claude Pricing in 2026: Plans, API Costs and How to Pay Less
By Adam, AI Engineer · 1 November 2026 · 4 min read
Claude costs $0 on the Free plan, $20 a month on Pro, $100 on Max and $200 on Ultra. If you are building a product, the API charges per token instead, from $0.10 per million input tokens on Haiku 5.5 to $10 on Fable 5.1. These are the prices on Anthropic's pricing page on 1 November 2026, before any tax.
Claude plans for people
| Plan | Price per month | Usage |
|---|---|---|
| Free | $0 | Limited, resets every 5 hours and every week |
| Pro | $20, or $17 billed yearly ($200 up front) | More than Free |
| Max | $100 | 5x Pro, plus $100 of API credits a month |
| Ultra | $200 | 10x Pro each week, plus $200 of API credits a month |
Pro includes access to the top Fable model through usage credits, while Max includes it for up to half of the weekly limit. Start with Pro unless you keep reaching its limits, because only then do Max and Ultra pay off.
Claude plans for teams
Team plans cover 2 to 150 seats. A Standard seat costs $25 a month, or $20 billed yearly. A Premium seat costs $125 a month, or $100 billed yearly, and gives 5x the usage of a Standard seat. Team plans also include up to $500 a month in pooled API credits.
Claude API prices
These are per million tokens, from the Claude API pricing documentation. The last column uses one example: 100,000 requests a month, each with 3,000 input tokens and 400 output tokens, and no caching.
| Model | Input | Cached input | Output | Example month |
|---|---|---|---|---|
| Haiku 5.5 | $0.10 | $0.01 | $0.50 | $50 |
| Sonnet 5.5 | $2 | $0.10 | $10 | $1,000 |
| Opus 5.5 | $4 | $0.20 | $20 | $2,000 |
| Fable 5.1 | $10 | $0.25 | $50 | $5,000 |
Haiku 5.5 rises to $0.50 input and $2.50 output for prompts over 100,000 tokens.
How to pay less
- Prompt caching: if most of each prompt repeats, reading it from the cache costs far less. With 70% of the input cached, the Sonnet 5.5 example falls from $1,000 to about $601, before cache write fees. Writing to the 5-minute cache costs $2.50 per million tokens on Sonnet 5.5.
- Batch processing: delayed jobs get 50% off, so the same example costs $500. This suits reports and bulk work, not live chat.
- Use the cheapest model that passes your tests: the example costs 100 times more on Fable 5.1 than on Haiku 5.5, so test before paying for the top model.
Costs that are easy to miss
- Web search costs $10 per 1,000 searches, plus tokens for the results.
- US-only processing costs 1.1x the standard rate.
- Fast mode on Opus 5.5 costs 2x the standard rate.
- Token counts can change between models. Anthropic says the tokenizer in Claude 4.7 and later produces about 30% more tokens for the same text than older models, so measure your real usage when you compare.
Which option should you choose?
- You just use it yourself: try Free, then move to Pro if you keep hitting limits.
- Your team uses it every day: Team Standard seats give you shared billing and administration.
- You are building a product: use the API, because a subscription does not replace API billing. Max and Ultra only include a set amount of credits.
Compare Claude with GPT and Gemini on your own numbers in the free AI cost calculator, or read our GPT pricing and Gemini pricing guides.