Direct answer: OpenAI cut GPT-5.6 Luna API prices by 80% and GPT-5.6 Terra prices by 20% on July 30, 2026. Luna now costs $0.20 per 1 million input tokens and $1.20 per 1 million output tokens. Terra now costs $2 input and $12 output per 1 million tokens. GPT-5.6 Sol’s standard price remains $5 input and $30 output per 1 million tokens.
estimated cost = (input tokens ÷ 1,000,000 × input rate) + (output tokens ÷ 1,000,000 × output rate)
Example: 10 million input tokens plus 2 million output tokens now costs about $4.40 on Luna or $44 on Terra, before separate tool-call charges or long-context premiums.
New GPT-5.6 prices at a glance
| Model | Input / 1M | Cached input / 1M | Output / 1M | Change |
|---|---|---|---|---|
gpt-5.6-luna |
$0.20 | $0.02 | $1.20 | 80% lower |
gpt-5.6-terra |
$2.00 | $0.20 | $12.00 | 20% lower |
gpt-5.6-sol |
$5.00 | $0.50 | $30.00 | Standard price unchanged |
These are standard API token rates. OpenAI says tool-specific features can carry separate per-call fees. Its Luna and Terra model pages also state that prompts over 272,000 input tokens are charged at 2× the input rate and 1.5× the output rate for the full request. Cache writes cost 1.25× the uncached input rate.
Old vs new price: what you save
| Model | Old input / output | New input / output | Example: 10M input + 2M output |
|---|---|---|---|
| Luna | $1 / $6 | $0.20 / $1.20 | $22 before → $4.40 now |
| Terra | $2.50 / $15 | $2 / $12 | $55 before → $44 now |
| Sol | $5 / $30 | Unchanged | $110 |
The arithmetic uses token charges only. Your real invoice can also depend on caching, tools, processing mode, long prompts, retries and the number of output tokens generated.
GPT-5.6 cost calculator worksheet
Copy this into a spreadsheet and replace the workload figures:
Monthly input tokens: __________ Monthly output tokens: __________ Luna estimate = (input / 1,000,000 × 0.20) + (output / 1,000,000 × 1.20) Terra estimate = (input / 1,000,000 × 2.00) + (output / 1,000,000 × 12.00) Sol estimate = (input / 1,000,000 × 5.00) + (output / 1,000,000 × 30.00) Add separately: - tool-call charges - long-context premium, when applicable - cache-write charges - Fast mode premium, when used - retries and evaluation traffic
Worked monthly examples
| Workload | Luna | Terra | Sol standard |
|---|---|---|---|
| 1M input + 1M output | $1.40 | $14 | $35 |
| 10M input + 2M output | $4.40 | $44 | $110 |
| 100M input + 20M output | $44 | $440 | $1,100 |
These examples exclude cached-input discounts and all non-token fees, so use them for quick planning rather than invoice reconciliation.
Which GPT-5.6 model should you choose?
- Start with Luna for high-volume, cost-sensitive work: classification, extraction, routing, routine implementation, structured outputs and background agent steps. OpenAI describes Luna as its fastest and most affordable GPT-5.6 model.
- Test Terra for balanced everyday work: use it when Luna misses your quality threshold but the task does not justify Sol. OpenAI positions Terra as the balance between intelligence and cost.
- Reserve Sol for the hardest steps: complex planning, high-stakes reasoning and cases where evaluations show that the flagship model materially improves the outcome.
Do not switch production traffic based on price alone. Run the same representative evaluation set against each model, then compare task success, latency, retries, output length and human-review time. A cheaper token rate can still cost more per successful task if quality drops.
A practical migration checklist
- Export at least 20–50 representative production tasks and expected results.
- Record the current model’s success rate, latency, input/output tokens, retries and human-review time.
- Test Luna first on lower-risk, high-volume steps.
- Route failed or uncertain cases to Terra or Sol instead of using one model for everything.
- Use structured outputs and explicit acceptance criteria where possible.
- Track cache hits and check whether any prompt crosses the 272K long-context threshold.
- Review tool-call fees separately from token spend.
- Roll out gradually, with spend alerts and a rollback path.
What changed for ChatGPT, Codex and Fast mode?
OpenAI says ChatGPT and Codex subscription prices and quota budgets did not change. However, Terra and Luna now consume fewer credits in paid subscriptions. Terra and Luna remain available in ChatGPT Work, Codex and the OpenAI API; OpenAI says Free and Go users can access Terra, while Plus, Pro, Business and Enterprise users can choose Terra and Luna.
The company also introduced Fast mode in the API to replace Priority Processing. For GPT-5.6 Sol, OpenAI says Fast mode can deliver up to 2.5× faster speed than Standard processing at twice the price, without changing intelligence. Existing requests tagged priority automatically use Fast mode.
Important limitations
- The quoted calculations are estimates based on published token rates, not a guarantee of your final bill.
- Separate tool fees, processing modes, caching behavior and very long prompts can change total cost.
- AWS rollout began on July 30, according to OpenAI; cloud-provider timing and pricing can differ, so verify the platform you actually use.
- Model quality depends on the task. Benchmark your own workload before migrating.
FAQ
What is the new GPT-5.6 Luna price?
GPT-5.6 Luna costs $0.20 per million input tokens, $0.02 per million cached input tokens and $1.20 per million output tokens at the standard API rate.
What is the new GPT-5.6 Terra price?
GPT-5.6 Terra costs $2 per million input tokens, $0.20 per million cached input tokens and $12 per million output tokens at the standard API rate.
Did GPT-5.6 Sol get cheaper?
No. OpenAI says Sol’s standard API price remains unchanged. The new Sol option is Fast mode, which costs twice the standard rate and can run up to 2.5× faster.
Did ChatGPT subscription prices drop?
No. OpenAI says ChatGPT and Codex subscription prices and quota budgets remain unchanged. Terra and Luna usage now consumes fewer credits.
Should every workload move to Luna?
No. Luna is a strong first test for high-volume, cost-sensitive work, but teams should compare models on real tasks and route harder or higher-stakes work to Terra or Sol when evaluations justify it.
Official sources
- OpenAI: Advancing the price-performance frontier with GPT-5.6 — price cuts, availability, subscription-credit changes and Fast mode.
- OpenAI API docs: GPT-5.6 Luna — current price, context limit, long-context multiplier and supported features.
- OpenAI API docs: GPT-5.6 Terra — current price, context limit, long-context multiplier and supported features.
Last fact-checked: August 1, 2026. Pricing can change; verify OpenAI’s live documentation before committing a production budget.