GPT-6.1 Terra Pricing: Cost per Task vs Sol and Luna

Last updated: November 2026

Key takeaways

• Terra lists at $1 input and $5 output per 1M tokens, with cached input at $0.05.
• In our illustrative support-bot example, Terra costs about $300 a month, half of Sol and ten times Luna.
• Prompt caching moves the gap more than the model choice does when your prompts repeat.
• List price is only half the picture: price per task depends on how many tokens each task burns.

Axios recently reported that OpenAI's new model cuts prices again. That headline is the signal. The numbers below come from public price tables as of early November 2026 and should be checked against OpenAI's own pricing page before you build a forecast on them.

What is GPT-6.1 Terra pricing per 1M tokens?

The current standard-tier, short-context rates in the price tables we checked are:

| Model | Input $/1M | Cached input $/1M | Output $/1M | |---|---|---|---| | GPT-6 Astra | $10.00 | $1.00 | $50.00 | | GPT-6.1 Sol | $2.00 | $0.10 | $10.00 | | GPT-6.1 Terra | $1.00 | $0.05 | $5.00 | | GPT-6 Luna | $0.10 | $0.01 | $0.50 |

Terra sits exactly at half of Sol on every line. Luna is another ten times cheaper than Terra. Astra, the flagship, is ten times Sol on input. You can browse the full OpenAI price list on Calcaas to see how these rows sit next to older models.

How much does a support bot cost on Terra, Sol and Luna?

Say a support bot handles 50,000 conversations a month. Assume, purely for illustration, 3,000 input tokens and 600 output tokens per conversation. That is 150M input tokens and 30M output tokens a month.

| Model | Input cost | Output cost | Monthly total | Per conversation | |---|---|---|---|---| | GPT-6 Astra | $1,500 | $1,500 | $3,000 | $0.060 | | GPT-6.1 Sol | $300 | $300 | $600 | $0.012 | | GPT-6.1 Terra | $150 | $150 | $300 | $0.006 | | GPT-6 Luna | $15 | $15 | $30 | $0.0006 |

The spread from Luna to Astra is 100x for the same traffic. If your product charges $0.50 a conversation, Terra leaves a very different margin from Astra, and that is before you add retrieval, tool calls or retries.

Does prompt caching change the Terra vs Sol math?

Yes, and by a lot when prompts repeat. Suppose 2,000 of the 3,000 input tokens per conversation are a cached system prompt and knowledge snippets. That is 100M cached tokens and 50M fresh input tokens a month.

• Sol: 50M x $2 + 100M x $0.10 + 30M x $10 = $100 + $10 + $300 = $410.
• Terra: 50M x $1 + 100M x $0.05 + 30M x $5 = $50 + $5 + $150 = $205.

Caching cut Sol's bill by about 32% and Terra's by about 32% too, because the structure is the same. Notice what dominates now: output tokens. At $5 and $10 per million, output is 73% of the Terra bill and 73% of the Sol bill in this example. That is why shortening answers often beats shopping for a cheaper model. For a deeper look at how the older tier structure compared, see our GPT-6.1 Sol pricing breakdown.

Is Terra cheaper per task, or just per token?

Only per token is certain. Per task depends on whether the cheaper model needs more attempts, longer reasoning or extra tool calls. If Terra needs 1.5x the tokens to finish the same task, its advantage over Sol shrinks from 2x to roughly 1.3x. If it matches Sol's success rate, the full 2x holds. The safe method is to run 100 real tasks on each model, record tokens and success, and compute cost per successful task, not cost per call. Our earlier note on the GPT-5.6 price cuts and margin impact shows the same logic on the previous generation.

When should you pick Terra, Sol or Luna?

• Luna: classification, routing, extraction and short replies where quality is easy to verify.
• Terra: the default for customer-facing chat and drafting where Sol is overkill.
• Sol: harder reasoning and coding where a failed attempt costs more than the token difference.
• Astra: only where the task value justifies a 5x price over Sol.

A router that sends easy turns to Luna and hard turns to Terra or Sol usually beats any single model on margin. If you are sizing a chat product, the AI chatbot cost use case is a good starting template.

Takeaway: Terra halves Sol's list price, but caching, output length and retry rate decide your real cost per task.

Want to run your own numbers? Plug your token split into the Calcaas LLM cost calculator and compare all four tiers side by side.