API cost guide
Claude Fable 5.2 Pricing and API Cost Examples
A per-million-token price only becomes actionable after you know what one job sends and generates. We used Anthropic's published US dollar API rates for the following examples. Use them to estimate the cost of a brief review, then replace the sample counts with usage from your own requests.
The quick answer
Standard Claude Fable 5.2 costs $8 per million uncached input tokens and $40 per million output tokens on the Claude API. Fast mode costs $16 and $80 respectively. A standard request with 12,000 uncached input tokens and 2,000 output tokens costs $0.176 in this example.
Calculate a one-off review
Multiply each usage count by its rate and divide each subtotal by 1,000,000. For the example, input costs $0.096 and output costs $0.080. Include the entire billable output count returned by the API when reconciling the request. A short visible answer can follow a longer reasoning pass.
Price repeated reviews separately
Say a studio reviews ten draft descriptions against a shared brief. Reuse the stable prefix with the documented cache configuration and append the changing draft after it. On standard Fable 5.2, a five-minute cache write is $10 per million tokens and a hit is $0.20. Keep cache creation, cache reads and uncached input as separate entries from usage.
| Usage category | Rate per million tokens |
|---|---|
| Uncached input | $8 |
| Five-minute cache write | $10 |
| One-hour cache write | $16 |
| Cache hit | $0.20 |
| Output | $40 |
Compare the price of a completed task
At the same input and output counts, the one-off example costs $0.352 in Fast mode. Compare cost per accepted review when retry frequency differs. Record wrong quotes, rejected recommendations and repair requests so the cheapest attempt can be distinguished from the cheapest useful result.
Keep each bill in its own column
The table describes the first-party API. Claude app limits and credits follow the account's plan. Cloud partners may apply different rates and terms. If a brief moves to Ciyo, record generation and editing costs there separately using the quote or usage displayed in that workspace.
Start with three representative requests
Choose one short brief, one long brief and one revision with a reusable prefix. Record token fields, total cost and whether the output met the same acceptance rule. That sample gives you a range to budget from before you schedule a larger batch.
Work through a cache example before budgeting a batch
Illustrative example: a 10,000-token shared brief is written once to the five-minute cache and read by nine more requests before it expires. Its input charge is $0.100 for the write plus $0.018 for the reads, or $0.118. Sending that same brief uncached ten times would cost $0.800.
This comparison covers only the shared input prefix. Add each changing prompt, image input, output and any separately billed tools. If a request misses the cache or runs after expiry, use its actual usage categories rather than assuming the $0.118 example still applies.
Frequently asked questions
Is the $8 rate the price of a Claude subscription?
It is the standard API price for one million uncached input tokens. Use the Claude Help Center to check app-plan eligibility and allowance.
What happens if the cache misses?
The request's usage reflects what was processed as ordinary input and any cache write. Calculate from the returned fields instead of assigning the cache-hit rate to the whole prompt.
Budget the visual pass separately
Use Ciyo's image tools when an approved brief is ready for visual exploration. Check the generation quote for the settings and references you choose.