What does GPT-5.6 actually cost?
Tier prices, caching and real costs
Sol $5/$30, Terra $2.50/$15, Luna $1/$6 — but real cost depends on cache hit rates and tier strategy. Here are the full billing rules, reproducible cost examples and ways to save.
Official price table (per 1M tokens)
Cache writes cost 1.25x the input rate; cached reads get a 90% discount. All tiers share the 1.05M API context and 128K max output.
| Tier | Input | Output | Cache write (1.25x) | Cache read (−90%) |
|---|---|---|---|---|
| gpt-5.6-sol | $5 | $30 | $6.25 | $0.50 |
| gpt-5.6-terra | $2.50 | $15 | $3.125 | $0.25 |
| gpt-5.6-luna | $1 | $6 | $1.25 | $0.10 |
The GA caching upgrades are worth designing for
GPT-5.6 ships more predictable prompt caching at GA: explicit cache breakpoints (you decide where the prefix splits) and a 30-minute minimum cache life. For agent workflows that means pinning system prompts and tool definitions as the cached prefix can cut long-session input costs by up to 90%.
Three reproducible cost examples
Line-by-line at official rates — swap in your own volumes to check.
Everyday agent session (Terra)
800K input + 60K output: 0.8 x $2.50 + 0.06 x $15 = $2.00 + $0.90 = about $2.90. If 80% of the input hits cached reads ($0.25/M), input drops to 0.16 x $2.50 + 0.64 x $0.25 = $0.56, for a total of about $1.46.
Hard task (Sol, high cache hit)
500K input (400K of it cached reads) + 50K output: 0.1 x $5 + 0.4 x $0.50 + 0.05 x $30 = $0.50 + $0.20 + $1.50 = about $2.20. Caching brings a Sol deep-dive close to Terra's uncached price.
Bulk light tasks (Luna)
1,000 calls at 10K input + 1K output each (10M input + 1M output total): 10 x $1 + 1 x $6 = $16. The same volume on Sol runs $80 — pushing light tasks down to Luna is the most direct saving.
How the prices compare
Sol ($5/$30) matches Claude Opus 4.8 ($5/$25) on input and runs slightly higher on output; Terra ($2.50/$15) is exactly half of Sol yet beats the previous flagship GPT-5.5 on quality; Luna's input rate is one fifth of Sol's. In effect OpenAI has decoupled flagship capability from flagship price — most workloads no longer need to pay for Sol.
Prices and usage on QCode
QCode's /models page shows live rates matching OpenAI's official prices, billed by usage with no subscription gate. The three tiers share one API key and one quota with GPT-5.5, Claude and Gemini — low-latency access from mainland China with local payment.
Pricing FAQ
How exactly is caching billed?
Cache writes cost 1.25x the input rate (Sol $6.25, Terra $3.125, Luna $1.25 per 1M tokens); cached reads cost 10% of input (Sol $0.50, Terra $0.25, Luna $0.10). Since GA you also get explicit cache breakpoints and a 30-minute minimum cache life.
Is long context billed extra?
There is no long-context surcharge in the official price list: within the 1.05M API window everything bills at the table rates. Filling the window does grow input cost linearly, of course — the lever is caching your prefix, not truncating context.
Does switching tiers affect billing or quota?
No. The tiers are just different model ids, each billed at its own rate against the same quota. You can mix them per request in one project — say Sol to plan, Terra to execute, Luna for batch work.
Do QCode prices match OpenAI's official rates?
Yes. The /models page shows the three tiers at official rates (Sol $5/$30, Terra $2.50/$15, Luna $1/$6), billed by usage. QCode's value is the onramp: low latency from mainland China, local payment, and one key across Claude / GPT / Gemini.
Use GPT-5.6 at official rates
Sign up for QCode and use all three tiers pay-as-you-go — no subscription, no gate.