GPT-5.6 Luna API Pricing
$0.20 input · $1.20 output per 1M tokens — 'good enough' at a fraction of the price.
GPT-5.6 Luna charges $0.20 input and $1.20 output per million tokens — one-eleventh of Sol's output price — yet OpenAI's own benchmarks show it nearly matching GPT-5.5's peak performance at less than half the cost, and outperforming Claude Opus 4.8 on coding. The July 30 price cut (80% off) turned it from 'cheap' into 'almost free'.
Price per 1M tokens
GPT-5.6 Luna · USD official list price
| Tier | Rate |
|---|---|
| Input | $0.20 |
| Output | $1.20 |
| Cache hit | $0.02 |
DeepSeek lists official prices in CNY; USD uses ≈ 6.9 exchange rate
Best for
$0.20 / $1.20
Ultra-cheap input and output per million tokens.
80% price cut
The July 30, 2026 cut turned Luna into the budget champion.
$0.02 cached input
Cached input is effectively free at two cents per million tokens.
High-volume traffic with mostly uncached input: default interactive workhorses, bulk summarization, and budget-sensitive startups. Combined with DeepSeek V4 Flash off-peak, it forms the embarrassingly cheap 2026 stack.
Billing notes
Published API pricing as of September 2026, standard tier, USD per million tokens. Also qualifies for the 50% Batch API discount — Luna at $0.60 output per million tokens is nearly free for offline jobs.
Luna vs the competition
Luna's $1.20 output now loses to DeepSeek V4 Flash at every hour — Flash output is ¥8.00 peak ($1.16) and ¥4.00 off-peak ($0.58), both below $1.20, and its cached input economics are far cheaper. Luna only wins if you must stay entirely on OpenAI infrastructure.
Frequently Asked Questions
What is the price of GPT-5.6 Luna per million tokens?
GPT-5.6 Luna bills $0.20 per million input tokens and $1.20 per million output tokens, with cached input at just $0.02 per million. The July 30, 2026 price cut (80% off) turned it from 'cheap' into 'almost free'.
How good is Luna despite the low price?
OpenAI's own benchmarks show Luna nearly matching GPT-5.5's peak performance at less than half the cost, and outperforming Claude Opus 4.8 on coding. It's the definition of 'good enough, at a fraction of the price'.
When should I avoid Luna?
Luna's edge assumes mostly uncached input — its $0.20 input rate already counts as cheap. If your workload is output-heavy agent loops, a model with cheaper effective output per token may win; measure your real mix before locking anything in.
Related model pricing
Compare this model against the rest of the lineup.
Estimate your exact cost
Model your own workload — input/output tokens, cache hit rate, and peak-hour share — against any model mix with the interactive pricing calculator.
Open the Pricing CalculatorRecommended Tools We ARE USING
We are using these tools ourselves for Development / Deployment. Check out for more details.
Opencode Go
You will receive $5 when subscribed, which directly offsets the 1st month's Go fee.
Could be the cheapest option for calling DeepSeek V4 Flash model in the world.
Vultr
RReceive $300 to test out VVultr platform
Referred user must be active 30+ days and spend $10–$25
Railway
Deploy anything without the complexity
Connect your repo, Railway handles the rest.
腾讯云 Tencent Cloud
Cloud servers, cloud databases, COS, CDN, SMS and other cloud products are on special offer now.
AWS high-quality alternative, we use COS as a replacement for S3
硅基流动 SiliconFlow
¥16 universal voucher for all platforms, valid for 180 days from the date of receipt
A comprehensive product matrix supporting the full-process implementation of AI applications.
Warp
One of the best AI terminal tools in the world
Interestingly, we can directly use various top-tier models in Warp without leaving the terminal.