GPT-5.6 Pricing: Sol, Terra, and Luna Costs Explained

GPT-5.6 API pricing has three tiers, per million tokens: Sol costs $5 input and $30 output, Terra costs $2 input and $12 output, and Luna costs $0.20 input and $1.20 output. These are the current prices after OpenAI's 2026-07-30 cut.

OpenAI sells GPT-5.6 by the token, not by the seat. You pay one rate for the tokens you send in and a higher rate for the tokens the model writes back. Each tier has its own rate, so your bill depends on which tier you pick and how much text moves through it.

This page lists the current price for each tier, explains the July 30 price cut, and walks through real cost examples. You will also learn how input and output pricing differ, how prompt caching lowers the bill, and how these rates stack up against Claude and Gemini.

GPT-5.6 Pricing by Tier (Current Rates)

Here are the current GPT-5.6 API prices, per million tokens, after the 2026-07-30 cut. Sol is the flagship, Terra is the balanced everyday tier, and Luna is the fastest and cheapest.

Input tokens are the text you send to the model. Output tokens are the text it generates back. Output always costs more than input on every tier.

Sol — Input $5.00 / Output $30.00 per million tokens.

Terra — Input $2.00 / Output $12.00 per million tokens (cut from $2.50 / $15.00).

Luna — Input $0.20 / Output $1.20 per million tokens (cut from $1.00 / $6.00).

Pick the tier by the job. Sol handles complex coding, cybersecurity, science, and deep reasoning. Terra covers high-volume business work like support, internal tools, and document analysis. Luna is built for summarizing, drafting, classifying, and routine automation.

• Sol: $5 in / $30 out — hardest work, new max and ultra reasoning modes.
• Terra: $2 in / $12 out — balanced, everyday business volume.
• Luna: $0.20 in / $1.20 out — fastest and most affordable.
• All rates are per million tokens on the OpenAI API.

The 2026-07-30 Price Cut Explained

On 2026-07-30 OpenAI cut Luna by 80% and Terra by 20%, while Sol stayed the same. Luna's input dropped from $1.00 to $0.20 and its output from $6.00 to $1.20. Terra's input fell from $2.50 to $2.00 and its output from $15.00 to $12.00.

OpenAI framed the cut as pushing the price-performance frontier. It says Luna now delivers work comparable to models that were frontier-class a year ago, at roughly six cents on the dollar per task and nearly nine times the speed.

OpenAI also claims that on Agents' Last Exam, Luna outperforms Claude Fable 5 at an estimated cost per task nearly 99% lower. Treat vendor benchmark claims as marketing until you test on your own workload.

Most analysts read the cut as competitive pressure. Cheap frontier tiers like Gemini Flash and Claude Haiku, plus open-weight models, have squeezed the low end. For buyers, the takeaway is simple: routine, high-volume jobs on Luna just got dramatically cheaper.

• Luna: down 80% on both input and output.
• Terra: down 20% on both input and output.
• Sol: unchanged at $5 / $30.
• Driver: competition from cheap frontier models pushing prices down.

Worked Cost Examples for Each Tier

Processing 1 million input tokens plus 1 million output tokens costs $35 on Sol, $14 on Terra, and $1.40 on Luna. That single job shows how far apart the tiers sit.

The math is simple. Add the input rate times your input millions to the output rate times your output millions. For 1M in and 1M out: Sol is $5 + $30, Terra is $2 + $12, and Luna is $0.20 + $1.20.

Now a realistic monthly example. Say a support bot handles 2 million input tokens and 1 million output tokens per day. That is 60M input and 30M output per month.

On Luna that month costs about $48 (60 x $0.20 plus 30 x $1.20). On Terra it costs about $480 (60 x $2 plus 30 x $12). On Sol it costs about $1,200 (60 x $5 plus 30 x $30).

Same workload, wildly different bills. A support bot rarely needs Sol-grade reasoning, so running it on Luna instead of Sol here saves roughly $1,150 a month. Matching the tier to the task is where the real savings live.

• 1M in + 1M out: Sol $35, Terra $14, Luna $1.40.
• Support bot (60M in / 30M out per month): Luna ~$48, Terra ~$480, Sol ~$1,200.
• Wrong-tier tax: running that bot on Sol costs ~25x the Luna price.

Input vs Output Costs and Prompt Caching

Output tokens cost more than input tokens on every GPT-5.6 tier, usually around six times more. Sol charges $5 for input and $30 for output; Luna charges $0.20 for input and $1.20 for output. Generating text is the expensive part.

This matters for how you design prompts. Long inputs are relatively cheap, but long answers are not. If a job produces short answers, its output cost stays low even when you feed it a lot of context.

Prompt caching lowers the input cost when you reuse the same context. GPT-5.6 supports prompt cache breakpoints, so repeated system prompts or shared documents can be cached and billed at a reduced rate on later calls.

OpenAI has not published every cached-rate figure on this page, so verify the exact caching discount on the official pricing docs. The general idea holds: reused context should be cached, not resent at full price.

• Output is roughly 6x the input rate on each tier.
• Short answers keep output cost down even with big inputs.
• Prompt cache breakpoints cut the cost of reused context.
• Design prompts to cache shared context and limit answer length.

API Pricing vs ChatGPT Plan Pricing

API pricing and ChatGPT plan pricing are two separate things. The per-token rates on this page are for the OpenAI API, where you build your own apps and pay for exactly the tokens you use.

ChatGPT is the consumer product. It uses flat monthly plans instead of per-token billing, so a person can chat all day for one fixed fee. That model does not map to the API rates above.

Choose the API when you are wiring GPT-5.6 into software, automations, or products at scale. Choose a ChatGPT plan when people just need to use the assistant directly in a browser or app.

For a full breakdown of the developer rates, see our GPT-5.6 API pricing guide. To learn what you can use without paying, see our guide on whether GPT-5.6 is free.

• API: pay per token, build your own apps.
• ChatGPT: flat monthly plan, use the assistant directly.
• Scaling software = API. People chatting = ChatGPT plan.

How GPT-5.6 Pricing Compares to Claude and Gemini

GPT-5.6 now spans a very wide price range, from Sol at the premium end to Luna at the budget end, which lets it compete with both Claude and Gemini across tiers. Compare like tier to like tier, not flagship to budget.

At the budget end, Luna at $0.20 input and $1.20 output competes directly with Claude Haiku and Gemini's Flash tier. After the July 30 cut, Luna is one of the cheapest capable frontier options for routine work.

In the middle, Terra at $2 input and $12 output lines up against mainstream Claude and Gemini working tiers for support and document jobs. Sol at $5 and $30 sits in flagship territory alongside top Claude and Gemini models.

Prices shift often, so confirm current Claude and Gemini rates on their own pricing pages before you decide. The honest rule stays the same: match the tier to the task, then compare the same workload across vendors.

• Luna vs Claude Haiku and Gemini Flash — the budget lane.
• Terra vs mainstream Claude and Gemini working tiers.
• Sol vs flagship Claude and Gemini models.
• Always price the same job across vendors before choosing.