GPT Sol vs Terra vs Luna vs Astra: All Eight Models Compared

Sorting out GPT Sol vs Terra vs Luna vs Astra is harder than it should be: OpenAI's API lists eight models under four names, and the names do not follow the version numbers. Sol, Terra and Luna arrived together as the three tiers of GPT-5.6 on July 9, 2026. GPT-6 Astra followed on September 3, GPT-6 Sol and GPT-6 Luna on September 22, GPT-6.1 Sol on September 29 and GPT-6.1 Terra on October 26.^[1] There is no GPT-6 Terra, GPT-6.1 Luna or GPT-6.1 Astra. OpenAI's documentation returned "Not found" for all three model IDs on November 4, 2026.

Treat the names as tiers. The GPT-5.6 model pages say Sol "roughly corresponds to the unsuffixed model tier used in earlier GPT-5 families", Terra to the mini tier and Luna to the nano tier.^[2] Astra exists only in GPT-6 and sits above Sol. All eight share a 1,050,000-token context window, 922,000 maximum input tokens and 128,000 maximum output tokens, so price, reasoning settings and tool support separate them, not context length.^[3]

GPT Sol vs Terra vs Luna vs Astra at a glance

Standard-tier prices from each OpenAI model page, checked November 4, 2026.^[2][3] On every model in the table, a prompt with more than 272K input tokens is billed at 2x input and 1.5x output for the whole request. Reasoning tokens count as output tokens.^[4]

OpenAI's GPT-6 guide lists four models: Astra for the hardest work, GPT-6.1 Sol to balance cost and intelligence, GPT-6.1 Terra for fast visual and writing tasks, and GPT-6 Luna for cost-sensitive volume.^[5] GPT-6 Sol is still listed, but its page points readers to GPT-6.1 Sol as "the newer Sol model".^[3]

5.6 sol vs terra vs luna

Inside GPT-5.6 the three tiers have the same knowledge cutoff, the same context limits and the same reasoning-effort range, from none to max. The differences are price and the tier each name maps to. Terra costs half of Sol on input and 60% on output; Luna costs 5% of Sol on input and 6% on output.

The GPT-5.6 Sol rate needs a date next to it. OpenAI describes $4 input and $20 output as a 20% input cut and a 33% output cut, available "at least through November 21, 2026".^[2] Working backwards from those percentages, the earlier list price was about $5 and $30. Budget against the higher figure if your project runs past that date.

The built-in tools also differ in the Responses API. GPT-5.6 Terra and GPT-5.6 Luna list web search, file search, image generation and code interpreter. GPT-5.6 Sol lists web search and file search only.^[2] If your app generates images or runs code inside the model call, check that list before settling a gpt 5.6 sol vs terra choice on price.

Astra vs Sol

GPT-6.1 Sol costs one fifth of Astra per token: $2 against $10 for input, $10 against $50 for output. OpenAI describes it as "near-Astra performance" for complex coding, computer use and professional work.^[5]

The per-token ratio is not the per-task ratio. OpenAI says Astra "achieved stronger results using substantially fewer output tokens" in several evaluations, and that its estimated API cost per task was lower than earlier models despite the higher token price.^[5] Run a sample of your real tasks on both and compare total billed tokens before deciding which is cheaper for you.

GPT-6 Sol and GPT-6.1 Sol share a $2 / $10 list price, but they are not interchangeable:

• Cached input on GPT-6.1 Sol costs $0.10 per million tokens (5% of input), against $0.20 (10%) on GPT-6 Sol.^[3]
• GPT-6.1 Sol drops the none effort. OpenAI's migration note says to use low instead.^[5]
• GPT-6.1 Sol adds an Ultrafast mode at 6x the Standard price.^[3]

GPT-6 vs GPT-5.6 Sol

The newer Sol is the cheaper one. GPT-6.1 Sol lists at $2 / $10, while GPT-5.6 Sol is $4 / $20 even with its promotion. GPT-6.1 Sol also has a later knowledge cutoff: April 30, 2026 against February 16.^[3]

The catch is the request format. OpenAI supports both models on Chat Completions, but for GPT-6 Astra, GPT-6.1 Sol and GPT-6.1 Terra "tool calling requires Responses". GPT-6 Sol and GPT-6 Luna allow function calling on Chat Completions only with reasoning_effort set to none.^[5] An application that sends tools through Chat Completions to GPT-5.6 Sol cannot switch the model name to gpt-6.1-sol and keep working. It has to move to the Responses API first.

Is there a GPT-6 Terra?

No. Terra exists in GPT-5.6 and GPT-6.1 only. GPT-6.1 Terra is the one to compare against GPT-5.6 Terra, and it is cheaper: $1 / $5 against $2 / $12.^[3] OpenAI's migration advice for that move is specific: compare both models on representative images and workflows at the same reasoning effort, and move any Chat Completions tool calls to Responses.^[5]

Tool calling through Chat Completions

reAPI serves all eight models on its OpenAI-compatible Chat Completions endpoint only, so the tool rules above decide which models suit an agent that runs there. These are the results recorded in reAPI's model documentation:^[6]

For a Chat Completions agent that calls functions, the GPT-5.6 tiers and GPT-6 Sol or Luna are the models with recorded working function calls on reAPI. Use GPT-6.1 Sol there for single-shot reasoning, code review or writing without tools.

Choosing by workload

Our reading of OpenAI's descriptions and the limits above; test on your own tasks.

• Long, hard agent runs where a failed attempt costs more than tokens: GPT-6 Astra, measured per task.
• Most complex work without a Chat Completions tool loop: GPT-6.1 Sol.
• Product photos, screenshots, screen actions and drafting text: GPT-6.1 Terra, which OpenAI recommends for those jobs.^[5]
• Classification and extraction at volume: GPT-6 Luna. At $0.10 / $0.50 it costs half of GPT-5.6 Luna on input, less than half on output, and has a later cutoff.

A rough sense of scale for a request with 20,000 input tokens and 2,000 output tokens (reasoning included) at Standard rates: $0.30 on Astra, $0.12 on GPT-5.6 Sol, $0.06 on GPT-6 Sol or GPT-6.1 Sol, $0.03 on GPT-6.1 Terra and $0.003 on GPT-6 Luna. Real output counts differ by model and effort, so treat these as arithmetic, not measurements.

FAQ

What do Sol, Terra and Luna mean in GPT model names?

They are tiers. OpenAI maps GPT-5.6 Sol to the old unsuffixed tier, Terra to mini and Luna to nano.^[2] Astra is a fourth tier, above Sol, that only GPT-6 has.

What is the GPT-6 Astra context window?

1,050,000 tokens, with up to 922,000 input tokens and 128,000 output tokens per request.^[3] Prompts above 272K input tokens are billed at the long-context rate.

How many parameters does GPT-6 Luna have?

OpenAI does not publish parameter counts for any of these models. Its published specifications are context size, token limits, knowledge cutoff, modalities and price.^[3]

Calling Sol, Terra, Luna and Astra on reAPI

All eight models use one key and the same endpoint, POST /api/v1/chat/completions, with model IDs that match OpenAI's: gpt-6-astra, gpt-6.1-sol, gpt-6-sol, gpt-6.1-terra, gpt-6-luna, gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna.^[6] Standard rates on the model pages on November 4, 2026, per million tokens:^[7]

If you are still choosing between GPT Sol vs Terra vs Luna vs Astra, start with GPT-6.1 Sol or GPT-6.1 Terra, check whether your app needs tool calls on Chat Completions, and move to Astra or Luna only when your own task results justify the price difference. For deeper two-model comparisons, see GPT-6.1 Terra vs 6.1 Sol and GPT-6.1 Sol vs GPT-6 Astra.

References

1. OpenAI. Changelog. Retrieved November 2026 from developers.openai.com/api/docs/changelog
2. OpenAI. GPT-5.6 Sol, GPT-5.6 Terra and GPT-5.6 Luna model pages. Retrieved November 2026 from developers.openai.com/api/docs/models/gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna
3. OpenAI. GPT-6 Astra, GPT-6 Sol, GPT-6 Luna, GPT-6.1 Sol and GPT-6.1 Terra model pages. Retrieved November 2026 from developers.openai.com/api/docs/models
4. OpenAI. Reasoning models. Retrieved November 2026 from developers.openai.com/api/docs/guides/reasoning
5. OpenAI. Using GPT-6. Retrieved November 2026 from developers.openai.com/api/docs/guides/latest-model
6. reAPI. GPT model documentation (GPT-5.6, GPT-6 and GPT-6.1). Retrieved November 2026 from reapi.ai/docs/gpt-6-1-sol
7. reAPI. Model pages and pricing. Retrieved November 2026 from reapi.ai/models