Skip to the document
Madhuopen lab

AI Insights: Latency & Cost Optimization · part 14 of 17

rejected28 July 2026cost: API-only (n=10)

GPT-5.6-Luna: tested and disqualified

Best quality in the whole test — and still no role for it

The question

Does the newly available GPT-5.6-Luna deployment ('1/5th the price') change the model choice?

rejected

Best-in-test quality (8.0/7.75), but 1,926 hidden reasoning tokens push TTFT to 18 s and real cost 30% above GPT-5.1.

Quality8.0 / 7.75best in the entire test
TTFT18.1 sstreaming can't hide reasoning
Hidden reasoning1,926 tok median71% of the output bill
Real cost/answer$0.017330% MORE than GPT-5.1

The pricing illusion

Sticker price is $1/$6 per 1M tokens — nominally a fifth of GPT-5.1's output rate. But Luna is a reasoning model: a median 1,926 hidden reasoning tokens per answer are billed at the $6 output rate and never appear in the response. Those invisible tokens are 71% of the output bill, making the real cost $0.0173 per answer — 30% more than GPT-5.1 — while the user waits 18 seconds for a first word. Streaming can't bring that forward, because reasoning happens before any output token exists.

Results (n=10, V5 prompt)

Results (n=10, V5 prompt)
MetricGPT-5.6-LunaGPT-5.1 + V5Qwen W8
Opus / Sonnet judge8.0 / 7.758.0 / 7.57.95 / 7.0
TTFT18.1 s~1.6 s614 ms
Wall time22.2 s~15 s6.1 s
Real cost/answer$0.0173$0.0136$0.00176
GPT-5.6-Lunareasoning modelshidden tokenspricing