gpt-5.6-luna
List prices per 1M tokens, what provider caching gives back, and what the cheaper sibling costs for the same work.
$0.2input / 1M
$1.2output / 1M
$0.02cached input / 1M (90% off)
$0.25cache write / 1M · pays back after 2 reads
Above 272,000 input tokens: $0.4 in / $1.8 out / $0.04 cached.
Same workload, cheaper sibling
100k requests/month of 2,000 input (1,500 static) and 300 output tokens, at list prices. "With cache" assumes the static prefix is served from the provider cache.
| Model | Per request | Per month | With cache | vs gpt-5.6-luna |
|---|---|---|---|---|
| gpt-5.6-luna (this model) | $0.00076 | $76 | $49 | — |
List prices. Cheaper models need an eval before swapping; TokenZero cascade routing is opt-in and advisory by default; an opt-in shadow parity monitor is coming.
Use it through TokenZero
OPENAI_BASE_URL=https://tokenzero.sh/v1 # keep your own key; cache formatting on, routing advisory
Free up to $500/month of routed spend; paid plans are a share of verified savings (Pro Cap: never more than $199/mo). Pricing