Based on Anthropic’s published API pricing, 720k tokens with Fable 5 would cost $6.00 if all are input tokens , $36.00 if all are output tokens , or somewhere in between depending on your input/output split.

How the math works

Fable 5 pricing (Claude API, standard rates as of June 2026):

  • Input tokens: $10 per 1 million tokens
  • Output tokens: $50 per 1 million tokens

So for 720,000 tokens (0.72M) :

  • If all input :

0.72×$10=$7.200.72\times $10=$7.200.72×$10=$7.20

  • If all output :

0.72×$50=$36.000.72\times $50=$36.000.72×$50=$36.00

A more realistic mixed example: say 600k input + 120k output:

  • Input: 0.6×10=$6.000.6\times 10=$6.000.6×10=$6.00
  • Output: 0.12×50=$6.000.12\times 50=$6.000.12×50=$6.00
  • Total ≈ $12.00

调价关键点是：output tokens are 5× more expensive than input , so your cost is driven mostly by how many tokens Fable 5 generates, not how many you send in.

Quick cost estimator

If you know your rough split, plug into this:

Cost=(input_M×10)+(output_M×50)\text{Cost}=(\text{input\_M}\times 10)+(\text{output\_M}\times 50)Cost=(input_M×10)+(output_M×50)

where input_M and output_M are your token counts in millions. For 720k total, just adjust the split.

Ways to reduce effective cost

If you’re using the API heavily, these can materially lower your bill:

  • Prompt caching : cached input reads can drop to ~$1/1M tokens (with higher write prices), which helps if you reuse long contexts.
  • Batch API : roughly halves standard rates to about $5/1M input and $25/1M output.
  • Routing simpler tasks to cheaper models (e.g., Haiku/Sonnet-class) and reserving Fable 5 for complex, high-value work.

TL;DR

  • 720k tokens, all input: ~$7.20
  • 720k tokens, all output: ~$36.00
  • Typical mixed usage: often in the $8–$20 range for 720k tokens, depending on how much the model writes back.

Information gathered from public forums or data available on the internet and portrayed here.