Claude Fable 5.1 API Pricing & Cost Calculator
Claude Fable 5.1 costs $10/M input and $50/M output at standard Claude API rates. Its unusually low $0.25/M cache-read price makes repeated context the central cost decision; creating that cache still costs $12.50/M for five minutes or $20/M for one hour.
- Standard: $10 input · $50 output per 1M tokens.
- Prompt caching: $12.50/M five-minute writes · $20/M one-hour writes · $0.25/M cache reads.
- Batch: 50% off applicable token charges for asynchronous work.
- US-only inference: 1.1x applicable input and output token prices.
- API model identifier:
claude-fable-5-1.
Claude Fable 5.1 pricing table
| Usage | Standard | Batch |
|---|---|---|
| Uncached input | $10.00/M | $5.00/M |
| 5-minute cache write | $12.50/M | $6.25/M |
| 1-hour cache write | $20.00/M | $10.00/M |
| Cache read / refresh | $0.25/M | $0.125/M |
| Output | $50.00/M | $25.00/M |
Anthropic documents cache writes as 1.25x base input for five minutes and 2x for one hour. Its pricing documentation says Batch and prompt caching can be combined. Provider marketplaces may use different prices or billing units.
Claude Fable 5.1 cost calculator
Estimate token cost
This estimates model token charges only. Tools, retrieval, storage, marketplace markups and other services are excluded. It also assumes the token categories entered match the usage fields returned by your API response.
Why Fable 5.1 caching changes the calculation
Fable 5.1 keeps the $10/M input and $50/M output structure but lowers cache reads to $0.25/M. A cache only saves money after reuse: writing 100K tokens to a five-minute cache costs $1.25, while reading those 100K tokens later costs $0.025 instead of $1.00 as ordinary input.
Practical examples
| Workload | Estimated cost | What drives it |
|---|---|---|
| 100K uncached input + 20K output | $2.00 | $1 input + $1 output |
| 100K five-minute cache write + 20K output | $2.25 | $1.25 cache creation + $1 output |
| 100K cache read + 20K output | $1.025 | $0.025 cache read + $1 output |
| Batch: 100K input + 20K output | $1.00 | 50% Batch discount |
Fable 5.1 versus GPT-6 Astra cost mechanics
Both publish $10/M standard input and $50/M output, but their calculators should not be treated as interchangeable. Fable distinguishes five-minute and one-hour prompt-cache writes, offers $0.25/M cache reads and has an optional US-only inference multiplier. GPT-6 Astra instead has a pricing threshold above 272K input tokens. Compare the request pattern, not just the headline rate.
Frequently asked questions
Does Fable 5.1 have a long-context pricing threshold?
Anthropic's current Fable 5.1 launch and pricing material does not document a separate long-context surcharge for this model. This calculator therefore applies no inferred threshold.
Is the Claude app subscription included?
No. This page covers usage-based API pricing, not Pro, Max, Team or Enterprise seat pricing.
Can safeguards route a request to another model?
Yes. Anthropic says some flagged cybersecurity or biology API requests may use configured fallback behavior. The company says customers are not charged Fable prices for rerouted requests; actual billing should be checked in API usage data.
Official sources
- Anthropic Claude Fable 5.1 announcement and pricing
- Anthropic Claude API pricing and prompt-caching multipliers