$1 / $6
Luna per million tokens in / out, sharing the family's 1.05M context window
Blog
GPT-5.6 Luna pricing: the volume-tier bake-off
Luna is OpenAI's volume tier at $1 input / $6 output per million tokens: the price where classification, extraction, and short summaries live. Its direct rival is Claude Haiku 4.5 at $1/$5, and at this end of the ladder a dollar per million is a real margin.
Skip the read: get a measured recommendation in a few quick questions. Run the Stack Finder
$1 / $5
Claude Haiku 4.5, the direct rival: tied input, 17% cheaper output
$0.0016
our measured Haiku bill for a real invoice-extraction task. Volume-tier work is this cheap done right
GPT-5.6 Luna costs $1 per million input tokens and $6 per million output as of July 2026, OpenAI's volume tier. Its closest rival, Claude Haiku 4.5, ties on input and charges $5 on output, 17% less. Both families discount cache reads to a tenth of the input rate.
How much does GPT-5.6 Luna cost?
$1 per million input tokens, $6 per million output (OpenAI pricing reference), with the family's caching structure: reads at 10% of input, writes at 1.25x. At these prices the discounts are fractions of fractions of a cent, which is the point: this is the tier where you stop thinking per request and start thinking per million requests.
Want this priced against your own workload? Run the Stack Finder or start with $100 in credits.
Luna vs Haiku 4.5: the bake-off that matters
At volume, the $6-vs-$5 output gap is a 17% unit-cost difference on the heavy side of the meter. Applied to a short extraction task's token profile, Luna prices at roughly $0.0019 per task on list math against the $0.0016 we metered on Haiku 4.5. Per million such tasks: about $1,900 versus $1,600. The honest caveat runs both directions: at this tier, quality cliffs cost more than rates do, and each family fails differently. Run both on your worst hundred cases before believing either rate card, and remember a volume-tier failure that escalates to a frontier retry costs 20x the task you tried to save on.
The same two tasks on every model we could meter
We sent an identical dashboard-generation prompt and an identical invoice-extraction prompt to every model we can call directly, on July 28, 2026, and recorded the actual bills. No estimates in the measured rows.
| Model | Rate $/M in / out | Dashboard task, measured | Extraction task, measured |
|---|---|---|---|
| Claude Fable 5 | $10 / $50 | $0.1462 | $0.0148* |
| Claude Opus 5 | $5 / $25 | $0.1158 | $0.0114 |
| Claude Opus 4.8 | $5 / $25 | $0.0724 | $0.0093 |
| Claude Sonnet 5 | $2 / $10 | $0.0410 | $0.0037 |
| Pareto | at cost | $0.0196 | $0.0018 |
| Claude Haiku 4.5 | $1 / $5 | $0.0162 | $0.0016 |
Measured 2026-07-28, one shot each, default settings, list prices. *Fable extraction is same-tokens list math from our earlier run; every other cell is a metered bill. Prompts published verbatim in the Fable 5 breakdown. All outputs completed the task; quality-per-dollar comparisons need the benchmark receipts, not this table alone.
The volume tier's real job
Luna and Haiku exist because most production traffic is routine: classify this, extract that, summarize this thread. The craft is in the split: send the routine 80% to the volume tier, escalate the hard 20%, and never pay frontier rates for boilerplate. Making that split per request, with receipts, is the entire product we sell: which is exactly why this page tells you the volume tiers are genuinely good. They are what good routing routes to.
Luna pricing questions, answered
Is Luna or Haiku 4.5 cheaper?
Haiku, by 17% on output ($5 vs $6 per million) with input tied at $1. On short structured tasks that is roughly $0.0016 versus $0.0019 per task. At a million tasks a month, a $300 monthly difference, if quality is equal on your traffic, which is the thing to verify.
Is Luna good enough for coding?
For boilerplate, snippets, and simple transformations, frequently. Our measured runs show even Haiku producing a working one-shot dashboard. For multi-file agentic work, failure-and-retry costs erase the savings fast; that traffic belongs a tier or two up.
What does Luna's 1.05M context window mean at this price?
That very long inputs are affordable to read: a full 1M-token context load costs about $1 uncached, $0.10 cached. It does not mean long context is free to reason over; quality on long-context tasks is the thing to test, not the price.
When does the volume tier become a false economy?
When failure rates climb: every failed cheap run that escalates to a frontier retry bills you twice, and review time is a cost too. Measure cost per successful task, not cost per request; the gap between those two numbers is where volume tiers quietly lose.
Not sure which model fits?
The Stack Finder asks a few quick questions about your workload and gives you a straight recommendation. No account required.
The volume tier, picked for you request by request. $100 to verify.
Buy Pareto API credits