What Does GPT-5.6 Luna Cost?
GPT-5.6 Luna costs $0.20 per million input tokens and $1.20 per million output tokens. That is the cheapest tier in OpenAI's current line.
Every figure on this page was checked on 16 September 2026 against the vendors' own published pricing. All prices are USD per million tokens, input first.
Two things follow from that pair of numbers. A blended million in plus a million out costs $1.40. And output costs 6x what input costs.
That second ratio decides more invoices than people expect. Once your output volume passes roughly one sixth of your input volume, output tokens become the majority of your Luna bill. At a 5:1 input to output ratio, output is already about 55% of the total.
AI Perks tracks $7.7M in AI credits across 194 companies, which is where a bill like this is supposed to go.
Luna Against Every Current OpenAI Tier
Luna is exactly 10x cheaper than gpt-5.6-terra and about 43x cheaper than gpt-6-astra on a blended million in, million out basis.
| OpenAI model | Input / 1M | Output / 1M | 1M in + 1M out | vs Luna |
|---|---|---|---|---|
| gpt-6-astra (flagship) | $10.00 | $50.00 | $60.00 | 42.9x |
| gpt-6-astra above 200k context | $20.00 | $75.00 | $95.00 | 67.9x |
| gpt-5.5 | $5.00 | $30.00 | $35.00 | 25.0x |
| gpt-5.6-sol | $4.00 | $20.00 | $24.00 | 17.1x |
| gpt-5.4 | $2.50 | $15.00 | $17.50 | 12.5x |
| gpt-5.6-terra | $2.00 | $12.00 | $14.00 | 10.0x |
| gpt-5.4-mini | $0.75 | $4.50 | $5.25 | 3.75x |
| gpt-5.4-nano | $0.20 | $1.25 | $1.45 | 1.04x |
| gpt-5.6-luna | $0.20 | $1.20 | $1.40 | 1.0x |
The gpt-5.4-nano row is the one to stare at. It matches Luna exactly on input and costs $0.05 more per million output tokens, about 3.6% more on the blended figure. Between those two, price is not the deciding factor. Pick whichever handles your prompts better.
A naming note, because it confuses people reading two price lists side by side. The Sol, Terra and Luna names replaced the older Instant, Thinking, mini and nano scheme in July 2026. The gpt-5.4 family kept the old convention, which is why both appear in the same current price list.
Long context is a separate line item on the flagship: above 200,000 tokens, astra jumps to $20.00 and $75.00. OpenAI has not published a long-context surcharge for Luna that we were able to verify, so assume the flat rate and check the console before committing to a long-context design.
Luna Against Claude and Meta Muse
Luna undercuts Claude Haiku 4.5, Anthropic's cheapest current model, by about 4.3x blended, and Meta Muse Spark by about 3.9x.
| Model | Input / 1M | Output / 1M | 1M in + 1M out | vs Luna |
|---|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $50.00 | $60.00 | 42.9x |
| Claude Mythos 5.1 (limited) | $10.00 | $50.00 | $60.00 | 42.9x |
| Claude Opus 5 | $5.00 | $25.00 | $30.00 | 21.4x |
| Claude Sonnet 5 | $2.00 | $10.00 | $12.00 | 8.6x |
| Claude Haiku 4.5 | $1.00 | $5.00 | $6.00 | 4.3x |
| Meta Muse Spark 1.1 to 1.3 | $1.25 | $4.25 | $5.50 | 3.9x |
| Meta Muse Spark Contributor | $0.10 | $0.20 | $0.30 | 0.21x |
| gpt-5.6-luna | $0.20 | $1.20 | $1.40 | 1.0x |
Here is the part almost nobody reported. Anthropic had scheduled Claude Sonnet 5 to rise from $2/$10 to $3/$15 on 1 September 2026, and then cancelled the increase. Anthropic's pricing page states: "The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur." The $2/$10 rate, originally announced as introductory through 31 August 2026, is now simply the standard price. Comparison posts written in July still quote the increase as though it took effect.
Meta's Contributor tier is the only current model on this page priced below Luna, and it is cheap for a stated reason: Meta trains on your prompts and completions on that tier. Meta also runs no startup credit program for Muse, so there is no credit route to offset it. AI Perks covers the providers that do run one.
On xAI and Grok, published token prices vary by model and third-party sources conflict badly enough that printing a number here would be guessing. Check xAI's own console. Google Gemini is left out of these tables entirely because we did not verify Gemini pricing in this pass, and an invented price is worse than a missing one.
What Luna Costs at Real Monthly Volume
At 50 million input and 10 million output tokens a month, GPT-5.6 Luna costs an estimated $22.00. The same workload runs $220 on terra, $400 on sol and $1,000 on gpt-6-astra.
These are estimates derived from the list prices above, at a 5:1 input to output ratio, before any caching, batching or volume terms.
| Model | 50M in | 10M out | Estimated monthly |
|---|---|---|---|
| gpt-5.6-luna | $10.00 | $12.00 | $22.00 |
| gpt-5.4-nano | $10.00 | $12.50 | $22.50 |
| gpt-5.4-mini | $37.50 | $45.00 | $82.50 |
| Claude Haiku 4.5 | $50.00 | $50.00 | $100.00 |
| Meta Muse Spark | $62.50 | $42.50 | $105.00 |
| Claude Sonnet 5 | $100.00 | $100.00 | $200.00 |
| gpt-5.6-terra | $100.00 | $120.00 | $220.00 |
| gpt-5.6-sol | $200.00 | $200.00 | $400.00 |
| Claude Opus 5 | $250.00 | $250.00 | $500.00 |
| gpt-6-astra | $500.00 | $500.00 | $1,000.00 |
Scale that tenfold and the gap gets serious: 500M in and 100M out is an estimated $220 a month on Luna and $10,000 a month on astra.
Credits are what close that gap in year one. $1,000 of OpenAI API credit covers roughly 45 months of the Luna workload above, or about 5 billion input tokens if you run input only. The same $1,000 buys one month on the flagship. AI Perks tracks what is currently on offer across 194 companies.
Which OpenAI Tier Should You Actually Pick?
Route by output volume and error tolerance. Price alone points at Luna for anything high volume with short outputs.
We have not benchmarked these models, and no capability ranking on this page is claimed or implied. What follows is cost reasoning only.
Luna at $0.20/$1.20 suits classification, extraction, routing, tagging, moderation, and any job where the prompt is long and the answer is short. At 6x output cost, the shape that saves you money is a lot of input and a little output.
Terra at $2.00/$12.00 is the 10x step up. Worth it only when Luna's output fails often enough that your retry rate plus your review time exceeds the price difference. Measure that, do not assume it.
Sol at $4.00/$20.00 sits 17x above Luna, and astra at $10.00/$50.00 sits 43x above it. At those multiples, the honest test is whether the task would fail without them.
The pattern that works: run the cheap tier by default, escalate on a confidence signal, and keep the flagship for the small share of traffic that genuinely needs it. A mixed router costs far less than paying flagship rates on every call. AI Perks is how founders cover what is left.
The Legacy Tiers Are Cheaper, and They Shut Down on 10 December 2026
gpt-5-nano at $0.05/$0.40 is the cheapest input price on this page. Its API shuts down on 10 December 2026, so do not build on it.
| Legacy model | Input / 1M | Output / 1M | 1M in + 1M out | Status |
|---|---|---|---|---|
| gpt-5 | $1.25 | $10.00 | $11.25 | API shutdown 10 Dec 2026 |
| gpt-5-mini | $0.25 | $2.00 | $2.25 | API shutdown 10 Dec 2026 |
| gpt-5-nano | $0.05 | $0.40 | $0.45 | API shutdown 10 Dec 2026 |
Two useful comparisons come out of that table. Luna is 38% cheaper than gpt-5-mini, the legacy tier it most closely replaces. And gpt-5-nano is about 3.1x cheaper than Luna, which is the real reason teams keep quoting it, on a model with a published end date.
If a pricing article you are reading still leads with $0.05 per million input tokens, it was written before the current line existed. Anything shipping past December belongs on Luna or gpt-5.4-nano, not on gpt-5-nano.
Frequently Asked Questions
How much does GPT-5.6 Luna cost per million tokens?
GPT-5.6 Luna costs $0.20 per million input tokens and $1.20 per million output tokens, verified on 16 September 2026. A blended million in plus a million out comes to $1.40. Output costs 6x input, so output length drives the bill on anything high volume. Credits to cover it are tracked at getaiperks.com.
Is GPT-5.6 Luna the cheapest OpenAI model?
It is the cheapest tier in the current line. One legacy model prices lower, gpt-5-nano at $0.05/$0.40, but its API shuts down on 10 December 2026. Among models OpenAI still supports, Luna at $0.20/$1.20 edges out gpt-5.4-nano at $0.20/$1.25.
How much cheaper is Luna than GPT-6 Astra?
About 42.9x on a blended million in, million out basis: $1.40 against $60.00. At 50 million input and 10 million output tokens a month, that is an estimated $22 against $1,000. Above 200,000 tokens of context, astra rises to $20.00/$75.00, widening the gap to roughly 68x.
Is Luna cheaper than Claude Haiku 4.5?
Yes, by about 4.3x blended. Haiku 4.5 costs $1.00/$5.00 per million tokens against Luna at $0.20/$1.20. Claude Sonnet 5 at $2.00/$10.00 is roughly 8.6x Luna, and Claude Opus 5 at $5.00/$25.00 is about 21.4x. Credits for both providers are listed at getaiperks.com.
Did Anthropic raise Claude Sonnet 5 pricing in September 2026?
No. The scheduled rise from $2/$10 to $3/$15 on 1 September 2026 was cancelled. Anthropic's page states the increase "will not occur," and the rate first announced as introductory through 31 August 2026 is now the standard price. Plenty of comparison articles still show the higher number.
How do I get free OpenAI API credits in 2026?
Several providers run credit programs for startups and developers, and the amounts vary widely by provider and profile. AI Perks tracks $7.7M in credits across 194 companies in one place, including current terms, so you are not checking a dozen separate provider pages every quarter.
Pick the cheap tier. Then get someone else to pay for it.