AI cost calculator

Claude Code cost calculator

What Claude Code costs at API prices for your model and workload, per session, per day and per developer per month, and how that compares with a flat Pro or Max plan. Anthropic reports an average of about $13 per developer per active day across enterprise deployments.

Your inputs

Price per million input / output tokens.
A session is one task from prompt to finished change, about the size of the example below.
tokens
Defaults are the example session in Anthropic's Claude Code cost docs. Your own numbers are in /usage.
tokens
tokens
tokens
A 5-minute cache write costs 1.25x the input price, a 1-hour write 2x.
Claude 4.7 and later models produce about 30% more tokens for the same text. The default example was measured on Sonnet 4.6.

Result

$150 per developer per month on Claude Sonnet 5.5

$7.13 per active day. At API prices this is more than the Pro and Max 5x plan prices, if the plan's usage limits cover your work.

Cost per session
$0.36
Cost per active day
$7.13Anthropic's enterprise average is about $13; 90% of users stay under $30
Per developer per month
$150
Team per month (1 developer)
$150
Team per year
$1,798

Worked examples

Each row is the calculator run on the inputs shown, with every other input at its default.

WorkloadModelSessions a dayPer dayPer developer per month
Light: a few fixes a dayClaude Sonnet 5.55$1.78$37.46
Typical dayClaude Sonnet 5.520$7.13$150
Heavy: parallel agentsClaude Sonnet 5.560$21.40$449
Typical day on Opus 5.5Claude Opus 5.520$14.27$300
Heavy on Opus 5.5Claude Opus 5.560$42.81$899
Typical day on Haiku 5.5Claude Haiku 5.520$2.39$50.29
Typical day on Fable 5.1Claude Fable 5.120$29.56$621

How it is calculated

Claude Code is billed by tokens. Each session reads most of its context from the prompt cache, writes new context to the cache, sends a little uncached input and receives output. The cost of one session is:

session cost = (cache reads x cache read price + cache writes x cache write price + input x input price + output x output price) / 1,000,000 x 1.3 if your counts come from a Claude 4.6 or older model (newer tokenizer) x 1.1 with US-only inference day = session cost x sessions per day; month = day x active days

Prices are Anthropic's list prices per million tokens. A 5-minute cache write costs 1.25 times the input price and a 1-hour write 2 times; a cache read costs 0.1 times (0.025 times on Fable 5.1). Haiku 5.5 has a higher price for prompts over 100,000 tokens, which the calculator applies when a session's context is that large.

The default session is the example Anthropic shows for the /usage command: 940,000 cache read tokens, 50,000 cache write tokens, 1,200 input and 5,300 output tokens. Run /usage in Claude Code after a few real sessions and enter your own numbers for a better estimate.

Pro and Max plans include Claude Code for a flat monthly price, with usage limits that reset every five hours plus weekly limits. Anthropic does not publish those limits as a token count, so this page compares API cost with the plan price only; it does not say how many sessions a plan allows.

Questions

How much does Claude Code cost per month?

On a subscription it costs the plan price: Pro is $20 a month ($17 a month billed yearly) and Max is $100 or $200 a month, with usage limits. On API pay-as-you-go it depends on model and workload; Anthropic reports an average of $150 to $250 per developer per month across enterprise deployments.

How much does Claude Code cost per day?

Anthropic reports an average of about $13 per developer per active day at API prices, and under $30 a day for 90% of users. With the default example session on Sonnet 5.5, 20 sessions a day comes to the figure shown above.

Is Claude Code cheaper on the Max plan or on the API?

If your API cost per month is above the Max price and your work fits inside the plan limits, the plan is cheaper. The chart above puts both on one scale for your inputs.

Why are cache reads the biggest number?

Claude Code sends the conversation and project context with every request, and most of it is served from the prompt cache at a tenth of the input price. Cache reads are cheap per token but there are many of them.

Which model is cheapest for Claude Code?

Haiku 5.5 has the lowest price per token, then Sonnet 5.5, Opus 5.5 and Fable 5.1. Cheaper models can need more turns for hard tasks, so compare cost per finished task, not only per token.

Sources

Prices and plans change. This page was last checked on 2026-10-29. The data is open: llm-prices.json (CC BY 4.0).

Related calculators