GPT-5.6 Luna vs GPT-5.6 Terra
GPT-5.6 Luna is the cheaper of the two, at 10x less on a blended 3:1 input-to-output rate. GPT-5.6 Luna runs $0.20 in / $1.20 out per 1M tokens; GPT-5.6 Terra runs $2.00 in / $12.00 out.
On the heaviest scenario below, that gap is $4,680 a month ($56,160 a year) for identical volume. Verified 2026-08-16.
Side by side
| GPT-5.6 Luna | GPT-5.6 Terra | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Input per 1M tokens | $0.20 | $2.00 |
| Output per 1M tokens | $1.20 | $12.00 |
| Cached input per 1M | $0.02 | $0.20 |
| Batch discount | Not offered | Not offered |
| Output-to-input ratio | 6.0x | 6.0x |
| Blended per 1M (3:1) | $0.450 | $4.50 |
What each costs on the same workload
Rate cards are hard to compare directly because the input:output mix changes the answer. These four scenarios hold the workload fixed and vary only the model, with no caching or batch discount applied to either side.
| Workload | GPT-5.6 Luna | GPT-5.6 Terra | Monthly difference | Cheaper |
|---|---|---|---|---|
| Support chatbot 50,000 conversations/month at 3K input and 500 output tokens each | $60.00 | $600.00 | $540.00 | GPT-5.6 Luna |
| RAG document search 200,000 queries/month at 6K retrieved-context input and 400 output tokens | $336.00 | $3,360 | $3,024 | GPT-5.6 Luna |
| Coding agent 5,000 runs/month at 60K input and 8K output tokens per run | $108.00 | $1,080 | $972.00 | GPT-5.6 Luna |
| Bulk classification 5,000,000 items/month at 400 input and 20 output tokens each | $520.00 | $5,200 | $4,680 | GPT-5.6 Luna |
Which should you actually pick?
GPT-5.6 Luna is cheaper on both input and output, so on price alone it wins regardless of your token mix. That does not automatically make it the right call: these are different models with different capability profiles, and paying 10x more is justified whenever the more expensive model gets the task right on the first attempt and the cheaper one needs two or three tries, or needs human correction downstream.
Discounts can also flip the arithmetic. Prompt caching is the bigger lever of the two for anything that resends a stable prefix: see the prompt caching savings calculator.
GPT-5.6 Luna vs GPT-5.6 Terra: common questions
Is GPT-5.6 Luna or GPT-5.6 Terra cheaper?
GPT-5.6 Luna is cheaper overall, at 10x less on a blended 3:1 input-to-output rate ($0.450 vs $4.50 per 1M blended tokens). It is cheaper on both input and output.
What is the price difference between GPT-5.6 Luna and GPT-5.6 Terra on a real workload?
On the bulk classification scenario (5,000,000 items/month at 400 input and 20 output tokens each), GPT-5.6 Luna costs $520.00 a month and GPT-5.6 Terra costs $5,200. That is a difference of $4,680 a month, or $56,160 a year, for the same volume of work.
Does GPT-5.6 Luna or GPT-5.6 Terra have better discounts?
GPT-5.6 Luna offers cached input at $0.02. GPT-5.6 Terra offers cached input at $0.20. On cache-heavy or latency-tolerant workloads these can matter more than the headline rate.
Should I switch from GPT-5.6 Terra to GPT-5.6 Luna?
Only if the cheaper model actually does the job. Price is the easy half of the decision; the hard half is whether output quality holds on your task. Run both against a sample of real traffic, then use the model switching savings calculator to put a number on the annual difference before committing.