Unofficial guide
GPT-6 Astra API cost
API spend for GPT-6 Astra is metered in tokens. Your bill is roughly the sum of four lines — uncached input, cached input, cache writes, and output — then adjusted by service tier and (when applicable) the long-context cliff.
Token lines that drive cost
- Uncached input — prompt tokens that miss the prompt cache (highest input rate).
- Cached input — prompt tokens served from cache (much cheaper than uncached).
- Cache writes — tokens written into cache when you establish or refresh a cache entry.
- Output — completion tokens; usually the most expensive line per million.
Unofficial Standard card used on this site: input $10 / cached $1 / cache writes $12.50 / output $50 per 1M tokens.
Service tiers
- Standard — baseline rates.
- Batch / Flex — modeled here as 0.5× Standard (lower cost, different latency / scheduling).
- Fast — modeled as 2× applicable rates.
Use the live calculator to swap tiers and see per-request and monthly totals update instantly.
Subscription vs API
A ChatGPT subscription is not the same product as the GPT-6 Astra API. This page is only about metered API token pricing. Confirm current numbers on OpenAI’s GPT-6 Astra docs and API pricing.