llmrelay
// gpt

GPT-5.6 Terra

The mid-size GPT-5.6. Half the price of Sol, most of the capability.

GPT-5.6 Terra is the mid-tier model in the GPT-5.6 line - half the per-token cost of Sol, with most of the capability. It is OpenAI’s answer to Claude Sonnet 5: the everyday driver you reach for first, with the flagship held in reserve for when you actually need it.

Terra carries the same 1M+ context window as Sol and the same mature function-calling ecosystem. What you trade off is depth: on complex reasoning tasks Sol finds more, but on most production work Terra is fast enough and good enough.

We sell it at half OpenAI’s list price - $1.25 per million input tokens against $2.50. That makes it cheaper than Claude Opus 5 and only slightly more expensive than Claude Sonnet 5, while staying inside the OpenAI tool ecosystem.

Official list
llmrelay
Input / M tokens
$2.50
$1.25
Output / M tokens
$15.00
$7.50
Context window
—
1,050,000 tokens
Max output
—
128,000 tokens

What it costs you per month

Real-world budget scenarios. Numbers are simple sums — official list price vs llmrelay's 50% off tier.

Usage scenario
Official
llmrelay
Light coding (1M in / 200K out per month)
$5.50
$2.75save $2.75
Heavy Cursor / Cline user (50M in / 5M out per month)
$200.00
$100.00save $100.00
Production RAG (500M in / 20M out per month)
$1550.00
$775.00save $775.00

What it's good at

Best for

Pick GPT-5.6 Terra when

  • +You are building on OpenAI’s API and need a workhorse model. Terra is priced for production volume while keeping the same API shape and tool support as Sol.
  • +Agent loops on a budget. Terra costs half what Sol does per token, and most agent tasks do not need frontier reasoning at every step.
  • +High-volume chat, RAG, or API endpoints where latency and cost matter more than depth. Terra is fast and cheap enough to put in a request path.
  • +You need long context without paying flagship rates. The 1M window costs half what Sol charges for the same amount of input.

Choose something else when

  • !Breadth of review matters more than cost. On security audits, migration planning, or anything where missing a corner case is expensive, the flagship premium is worth it. Escalate to Sol or Claude Opus 5.
  • !You are optimising purely for cost. Claude Sonnet 5 is slightly cheaper at $1.00/M input, and GPT-5.6 Luna is a quarter of Terra’s rate if you can tolerate the capability drop.
  • !Paying mid-tier rates for ordinary copy. Fable is not a writing SKU. For hard Claude work use Opus 5 or Fable 5.1.
  • !Classification or routing at massive scale. GPT-5.6 Luna or Claude Haiku 4.5 are the right models for high-volume, shallow tasks.

Questions people ask about GPT-5.6 Terra

How much does GPT-5.6 Terra cost?

OpenAI’s list price is $2.50 per million input tokens and $15.00 per million output tokens. On llmrelay it is $1.25 and $7.50 - exactly half list. Billing is prepaid per token, with no subscription or volume commit.

Should I use GPT-5.6 Terra or Claude Sonnet 5?

Both are mid-tier models priced for volume. Terra costs $1.25/M input against Sonnet 5’s $1.00, so Sonnet is slightly cheaper. The choice is mostly about ecosystem: if your workflow is built around OpenAI function calling, Terra is the natural pick. If you are starting fresh or prefer Claude’s output style, Sonnet 5 is a strong alternative.

Is Terra good enough, or do I need Sol?

For most work Terra is enough. Start here and escalate to Sol on the specific tasks where you see Terra fall short. The 2× cost gap means starting on Sol and trying to trim down is more expensive than starting on Terra and escalating up.

What is the context window on GPT-5.6 Terra?

1,050,000 tokens input, with up to 128,000 output tokens. That is the same window as GPT-5.6 Sol, so long-context work does not require paying the flagship rate.

Try GPT-5.6 Terra at half the price

Free to create an account, no subscription, $10 minimum top-up — enough to run this model against a real task and judge quality yourself.

Get API key →

Compare GPT-5.6 Terra against the alternatives

The comparisons this model appears in, the models nearest it on price, and the full rate card.