GPT-5.6 Sol vs Terra vs Luna: Price, Context, and Which Model to Use
Compare GPT-5.6 Sol, Terra, and Luna API pricing, cache-write costs, long-context rates, model IDs, and practical routing choices for production workloads.
OpenAI’s GPT-5.6 release replaces the old flagship/mini/nano naming pattern with three durable tiers: Sol, Terra, and Luna. All three accept text and image input, support a 1.05M-token context window, and can return up to 128K output tokens. The main differences are capability, price, and the amount of work you should route to each tier.
Quick Comparison
Standard prices are per 1 million tokens.
| Model | API model ID | Input | Cached read | Cache write | Output | Best fit |
|---|---|---|---|---|---|---|
| GPT-5.6 Sol | gpt-5.6-sol | $5.00 | $0.50 | $6.25 | $30.00 | Hard coding, research, complex agents |
| GPT-5.6 Terra | gpt-5.6-terra | $2.50 | $0.25 | $3.125 | $15.00 | Default production model |
| GPT-5.6 Luna | gpt-5.6-luna | $1.00 | $0.10 | $1.25 | $6.00 | High-volume and cost-sensitive routing |
The unsuffixed gpt-5.6 alias routes to GPT-5.6 Sol. Use the explicit IDs when cost predictability matters.
Long-Context Pricing
When a request contains more than 272K input tokens, OpenAI applies the higher rate to the full request: 2x input pricing and 1.5x output pricing.
| Model | Long input | Long cached read | Long cache write | Long output |
|---|---|---|---|---|
| GPT-5.6 Sol | $10.00 | $1.00 | $12.50 | $45.00 |
| GPT-5.6 Terra | $5.00 | $0.50 | $6.25 | $22.50 |
| GPT-5.6 Luna | $2.00 | $0.20 | $2.50 | $9.00 |
Do not estimate a 500K-token repository request with the short-context row. Crossing the threshold changes the price of every token in that request, not only the tokens above 272K.
Example Monthly Cost
For a workload using 10M uncached input tokens and 2M output tokens per month, before tool-call fees:
| Model | Input cost | Output cost | Total |
|---|---|---|---|
| GPT-5.6 Sol | $50 | $60 | $110 |
| GPT-5.6 Terra | $25 | $30 | $55 |
| GPT-5.6 Luna | $10 | $12 | $22 |
Terra is exactly half Sol’s token price, while Luna is one fifth of Sol’s price. Actual task cost can differ because models may use different numbers of reasoning and output tokens to complete the same job.
Choose GPT-5.6 Sol When
- A failed attempt costs more than the model-price difference.
- The task involves difficult debugging, architecture, cybersecurity, science, or long-running agent work.
- You need OpenAI’s highest GPT-5.6 reasoning tier or want to use the
maxeffort setting. - You are evaluating
proreasoning mode. Pro is a mode on Sol, not a separategpt-5.6-proAPI model.
Choose GPT-5.6 Terra When
- You need a strong default for coding agents, document work, and production assistants.
- GPT-5.5-level workflows are too expensive on Sol but a small model is too risky.
- You want the same context and output limits as Sol at half the token price.
For most teams, Terra is the first model to benchmark as the default route.
Choose GPT-5.6 Luna When
- The workload is classification, extraction, summarization, search triage, or simple code changes.
- Requests are high-volume and can be escalated when confidence is low.
- You need a 1.05M context window but do not need Sol-level reasoning on every call.
Luna is not limited to tiny prompts. It has the same published context and maximum-output sizes as Sol and Terra, but quality should still be evaluated on your own tasks.
A Practical Routing Policy
- Start routine requests on Luna.
- Route normal coding and professional work to Terra.
- Escalate ambiguous, high-impact, or failed tasks to Sol.
- Send asynchronous traffic through Batch or Flex when latency is not important.
- Keep reusable prefixes stable so cached reads can receive the 90% discount.
Use the AI Model Pricing Calculator for your token mix, then compare the result with the full GPT-5.6 API Pricing Guide.