Cutoff Data

GPT-5.6 Luna: price, speed and score

Everything we have measured about GPT-5.6 Luna from OpenAI, taken from real billed API calls rather than a published rate card.

Last reading 8/7/2026, 5:06:15 PM UTC

Input / 1M

—

Output / 1M

—

Context

—

Median response

1.07s

p95 response

1.25s

Uptime, 7 days

100%

Benchmark score

87.8%

Cost per task

—

Checks, 7 days

Compare with

Gemini 3.6 Flash GPT-5.6 Terra GPT-5.6 Sol GPT-5.4 Mini GPT-5 Mini GPT-5 Nano Gemini 3.5 Flash Gemini 3.1 Flash Lite Gemini 3.1 Pro (preview)Gemini 2.5 Flash Gemini 2.5 Flash Lite Gemini 2.5 Pro

How this is measured

Every hour we send one identical prompt to GPT-5.6 Luna through a single gateway account, record how long the call took and what it was billed, then divide that cost by the tokens consumed to get an observed price per million tokens.

Benchmark figures come from a fixed suite of code-graded tasks re-run against every tracked model. Scores are the share of available points earned; cost is the average spend per task.

These are one buyer's readings on one route. Enterprise pricing, batch discounts and cached-input rates will differ.