Home → Guides → GPT-5.6 Terra in production

GPT-5.6 Terra in production

The mid-tier GPT-5.6, and in our own traffic the one that does most of the work for a fraction of the money. Numbers from 13 real billed calls.

The short version

Median cost per call was $0.01807 against $0.02333 mean, across 13 billed calls.
About a third of sol's cost per call in the same period, for work that mostly did not need sol.
68% of input arrived as cache reads — this is a workload shape that benefits heavily from a stable prompt prefix.
List price is $2.00 per 1M input, $12.00 per 1M output.

Measured against its siblings

Model	Calls	Avg cost / call	Avg first token	Avg output
`gpt-6-astra`	14	$0.1047	2,176 ms	372 tok
`gpt-5.6-sol`	29	$0.0796	11,873 ms	975 tok
`gpt-5.6-terra`	13	$0.0233	7,767 ms	124 tok

Different workloads rather than a controlled benchmark — average output length varies eightfold across those rows, which moves both cost and duration. Read it as a directional signal from real traffic, not a lab result.

What it actually cost

Per call	Cost at list price
Cheapest	$0.00774
Median	$0.01807
Mean	$0.02333
Most expensive	$0.06118

13 calls · 116,781 input tokens · 1,612 output · 252,160 cache-read · $0.30 total at list price.

Where it belongs

Terra is the sensible default for work that is neither trivial nor the final product: agent loops, bulk transformation, structured extraction that needs some judgement, drafts a human will edit.

Go down to gpt-5.6-luna ($0.49 / $2.01) for classification, routing, tagging and reformatting. Those dominate token counts and need no judgement.
Go up to gpt-5.6-sol ($5.00 / $30.00) only where output quality is what you are selling.

The spread from luna to sol is roughly ten times for the same API shape. Most pipelines that feel expensive are sending everything to one tier.

Calling it

from openai import OpenAI

client = OpenAI(api_key="YOUR_KEY", base_url="https://apiclan.us/v1")

resp = client.chat.completions.create(
    model="gpt-5.6-terra",
    messages=[{"role": "user", "content": "Hello"}],
)

Switching tiers is one word. Same endpoint, same key, same request shape — change gpt-5.6-terra to gpt-5.6-luna or gpt-5.6-sol and nothing else in your code moves. That makes per-stage routing cheap to try.

What you pay here

Per 1M tokens	List price	APICLAN	You save
Input	$2.00	$0.400	80%
Output	$12.00	$2.40	80%

Calculator and the rest of the family on the price page.

Questions people actually ask

Is terra good enough to replace sol?

For most non-final work, in our experience yes. But this is exactly the question you should not take from a vendor page — run both on twenty real inputs from your own product and read the outputs.

How do I route between tiers?

Split on deciding versus doing. Classification, extraction and routing go to luna; the work itself to terra; only the final user-visible answer to sol. Each is a one-word change.

Start using it

No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.

Read the 30-second quickstart

Prices quoted on this page are regenerated automatically from live billing data. Third-party terms are quoted from that party's own published documentation.