Live catalog rates, one honest unit: a sample article's worth of reading (1M in) plus a long draft (250K out), fee included. Price is not quality — test both against your prompt.
Model A
Claude Fable 5.1 · Anthropic
Rate / 1M
$0.01 in / $0.06 out
Sample (1M + 250K)
$27.00
Context
1M
Billing
Same credit pool
Model B · lower sample price
GPT-6 Luna · OpenAI
Rate / 1M
$0.00 in / $0.00 out
Sample (1M + 250K)
$0.27
Context
1M
Billing
Same credit pool
Three questions, in order
Cost to usable output — retries and quality move the real number, not just the unit price.
Fit for the input — files, context and modality before headline scores.
Workflow portability — the same balance moves to the next stronger model tomorrow.