Skip to main content

GPT-5.6 Terra

OpenAI

๐Ÿง  Reasoning๐ŸŒ Web Search

The balanced middle tier of the GPT-5.6 family. Most of Sol's ability at half the cost.

Released July 9, 2026

Pricing

Input tokens$2.50/M
Output tokens$15.00/M

Capacity

Context window1.1M tokens
Max output128K tokens

Capabilities

โœ“Reasoning & Planning
โœ“Web Search
โœ—Open Source

Best Scores

GDPval-AA v21593.0 pts
GPQA Diamond92.9%
Terminal-Bench 2.187.4%
SWE-bench Pro63.4%
Artificial Analysis Intelligence Index55.0 pts

Best For

๐Ÿ’ปcoding
๐Ÿ“Šanalysis
โœ๏ธwriting

Benchmark Scores

Specialized Skills

GDPval-AA v2

Real paid work from 44 different jobs, such as law, nursing, and software. Judges compare two answers side by side without knowing which model wrote them, and the winner gains rating points. A typical human expert scores 1000, so a higher number means the work was picked over a human more often.

1593.0 pts

Provider-reported; no independent run recorded yet.

Knowledge

GPQA Diamond

PhD-level science questions written so you cannot just Google the answer. Tests whether the model can reason about hard science.

92.9%

Provider-reported; no independent run recorded yet.

Humanity's Last Exam

PhD-level questions across many subjects. Tests deep reasoning on the hardest questions humans can ask.

41.8%

Independently measured by Artificial Analysis.

Software Engineering

Terminal-Bench 2.1

Puts the AI in front of a computer terminal and asks it to finish multi-step tasks on its own. Measures how good an "AI agent" it is.

87.4%

Provider-reported; an independent run by tbench.ai (Codex) lands at 78.4%.

SWE-bench Pro

The harder version of the coding test. Bigger codebases, trickier bugs. Scores drop for everyone, so the gaps between models become clearer.

63.4%

Provider-reported; no independent run recorded yet.

Reasoning

Artificial Analysis Intelligence Index

A frequently refreshed overall score made from nine modern tests: real work, tool use, terminal tasks, science, hard questions, and long-context reasoning. It is a scorecard rather than a percent correct.

55.0 pts

Independently measured by Artificial Analysis.

Why Choose GPT-5.6 Terra?

Terra is the Goldilocks option in the GPT-5.6 lineup. It delivers most of Sol's capability at half the cost, with a balanced feature set. Great for teams who want OpenAI's capabilities without premium pricing.

How It Compares

vs GPT-5.6 Sol

Terra is half the cost; Sol is faster and better at agentic work.

vs Claude Sonnet 5

At intro pricing, Sonnet is cheaper; Terra has broader capabilities.

You Might Also Like