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

In short: GPT-5.6 Sol is the most capable of the three GPT-5.6 models from OpenAI - Sol, Terra and Luna. Through OpenAI's API it costs $4 per million input tokens and $20 per million output tokens (a promotional price), with a 1.05-million-token context window. Terra is the mid-priced tier and Luna the cheapest; all three share the same context window, tools and settings.
What is GPT-5.6 Sol?
GPT-5.6 Sol is the flagship of OpenAI's GPT-5.6 family, which OpenAI describes as its "flagship model for complex professional work". People sometimes search for it as "ChatGPT 5.6 Sol": ChatGPT is OpenAI's chatbot app, while GPT-5.6 Sol is the name of the model itself, documented on OpenAI's model page for developers using the API.
OpenAI names its models in tiers, like sizes of the same product. In the GPT-5.6 family:
- Sol - the most capable and most expensive tier.
- Terra - the middle tier, balancing capability and cost.
- Luna - the cheapest tier, for high-volume, simple work.
GPT-5.6 is not one model with one price. OpenAI ships it as three named tiers - Sol, Terra and Luna - that share almost everything except capability and cost. OpenAI's own documentation maps them onto the naming earlier GPT-5 families used: Sol "roughly corresponds to the unsuffixed model tier", Terra to the mini tier and Luna to the nano tier. If you call the plain gpt-5.6 alias, your requests go to Sol.
This guide sets out what each tier costs, what they have in common, where the pricing has edges, and how GPT-5.6 relates to the GPT-6 family that followed it.
The three tiers at a glance
| Tier | Input / 1M | Cached input / 1M | Output / 1M | Earlier GPT-5 equivalent | LMArena |
|---|---|---|---|---|---|
| Sol | $4.00 | $0.40 | $20.00 | Unsuffixed (flagship) | 1485 Elo (rank 22, xhigh effort) |
| Terra | $2.00 | $0.20 | $12.00 | mini | 1467 Elo (rank 50, xhigh effort) |
| Luna | $0.20 | $0.02 | $1.20 | nano | 1453 Elo (rank 76, xhigh effort) |
Prices are OpenAI's published API rates for prompts up to 272K input tokens. Sol's are promotional: OpenAI states they are "available at least through November 21, 2026", and describes them as a 20% cut on input and a 33% cut on output. Terra and Luna carry no such note.
What all three share
The tiers are not a feature ladder. According to OpenAI's model pages, every one of them has:
- a 1,050,000-token context window and 128,000 maximum output tokens;
- a knowledge cutoff of February 16, 2026;
- the same reasoning effort settings -
none,low,medium(the default),high,xhighandmax; - text and image input, text output - no audio or video;
- the same endpoints, including Chat Completions, Responses, Realtime and Batch.
OpenAI's model pages rate reasoning and speed: Sol is rated "Highest" for reasoning, Terra "Higher" and Luna "High", with all three rated "Fast".
The pricing details that change the bill
The 272K line. Prompts with more than 272K input tokens are billed at 2x the input rate and 1.5x the output rate - for the whole request, not just the tokens over the line. A 300K-token prompt to Sol with a 10K-token answer costs $2.40 for input and $0.30 for output, where the same answer to a 270K prompt would cost about $1.08 and $0.20. A request that carries conversation history forward is billed at the higher rate once its input passes 272K tokens.
Caching. Cached input is a tenth of the normal input rate on every tier ($0.40, $0.20 and $0.02 per million). Writing to the cache is billed at 1.25x the uncached input rate.
Batch. OpenAI's Batch API gives a 50% discount on the synchronous rates, for work that can wait up to 24 hours for results.
How OpenAI describes each tier
| Tier | OpenAI's description | Reasoning rating | Speed rating | Price relative to Sol (input / output) |
|---|---|---|---|---|
| Sol | "flagship model for complex professional work" | Highest | Fast | - |
| Terra | "designed for workloads that balance intelligence and cost" | Higher | Fast | 50% / 60% |
| Luna | "designed for cost-sensitive, high-volume workloads" | High | Fast | 5% / 6% |
All three share the same context window, tools and reasoning settings, so a request moves between tiers by changing the model ID.
GPT-5.6 and the GPT-6 family
OpenAI released GPT-6 Astra on September 3, 2026, then GPT-6 Sol and GPT-6 Luna on September 22 and GPT-6.1 Sol on September 29. Two points about how the families relate:
- There is no GPT-6 Terra. The GPT-6 family has no mid tier; GPT-5.6 Terra remains OpenAI's model in that slot.
- LMArena ratings. On LMArena's text board GPT-5.6 Sol has 1485 Elo (rank 22, xhigh effort), GPT-6 Astra 1476 Elo (rank 37, max effort) and GPT-6 Sol 1458 Elo (rank 64, max effort). GPT-6 Sol costs half as much as GPT-5.6 Sol - $2 / $10 against $4 / $20 per million tokens. The figures side by side are on the GPT-6 Sol vs GPT-5.6 Sol page.
How it compares outside OpenAI
At $4 / $20, GPT-5.6 Sol is priced below both Claude Opus 5 ($5 / $25) and Claude Fable 5 ($10 / $50). Terra costs $2 / $12, against $2 / $10 for Claude Sonnet 5. Luna costs $0.20 / $1.20. Current Elo and benchmark figures for all of them are on the benchmarks page.
An earlier version of this guide quoted an Artificial Analysis Intelligence Index score of 61 for Sol. Artificial Analysis has since replaced that index version with one on a different scale, so the figure is no longer comparable and has been removed.
Sources
- OpenAI model pages: GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna
- OpenAI API pricing
- LMArena text leaderboard
Frequently asked questions
What is GPT-5.6 Sol?+
GPT-5.6 Sol is the most capable of OpenAI's three GPT-5.6 models, described by OpenAI as its flagship model for complex professional work. Through the API it costs $4 per million input tokens and $20 per million output tokens at its promotional price, with a 1.05-million-token context window.
Is GPT-5.6 Sol the same as ChatGPT?+
No. ChatGPT is OpenAI's chatbot app. GPT-5.6 Sol is the name of one of OpenAI's models, which developers can use through OpenAI's API.
What is the difference between GPT-5.6 Sol, Terra and Luna?+
Capability and price. All three have a 1.05-million-token context window, 128,000 maximum output tokens, the same tools and the same reasoning settings. Sol is the most capable and most expensive at $4 and $20 per million tokens, Terra is the middle tier at $2 and $12, and Luna is the cheapest at $0.20 and $1.20.
Which model does the gpt-5.6 alias use?+
GPT-5.6 Sol, according to OpenAI's documentation.
What is the difference between GPT-5.6 and GPT-6?+
GPT-6 is OpenAI's newer family, released in September 2026: GPT-6 Astra, GPT-6 Sol, GPT-6 Luna and GPT-6.1 Sol. GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, half of GPT-5.6 Sol's price. The GPT-6 family has no Terra tier.
Is there a GPT-6 Terra?+
No. The GPT-6 family released so far is Astra, Sol, Luna and GPT-6.1 Sol.
Does a long prompt cost more with GPT-5.6?+
Yes. Above 272,000 input tokens, the whole request is billed at twice the input rate and 1.5 times the output rate.
Related Guides
Enjoyed this? Get more in your inbox.
Up to ten of the week's AI news stories, by email once a week.