GPT-5.6 Terra is the balanced tier - 87.4% on Terminal-Bench 2.1 against Sol's 88.8%, 89.6% long-context recall against Sol's 91.5%, at roughly 500-900ms. The one most work should default to.
What ChatGPT 5.6 Terra is best for
Everyday production work where Sol is overkill
Interactive assistants and chat that must stay coherent turn to turn
Document analysis - 89.6% MRCR recall, within two points of the flagship
Anything you will run repeatedly, where latency and cost compound
How to use ChatGPT 5.6 Terra
Step 1
Make Terra your default and escalate deliberately. On most tasks you give up one to three benchmark points against Sol, which rarely justifies the extra wait.
Step 2
Paste whole documents rather than summarising them first - 89.6% recall means Terra handles long inputs properly, and that is the main thing it does that Luna cannot.
Step 3
Be explicit about format, length and tone. Terra follows stated instructions closely instead of padding answers out.
Example prompts
Frequently asked questions
What is GPT-5.6 Terra?
Terra is the middle model of the GPT-5.6 family. On most published benchmarks it lands one to three points behind Sol - 87.4% vs 88.8% on Terminal-Bench 2.1, 77.4 vs 80 on the Artificial Analysis Coding Agent Index.
When is Terra enough, and when do I need Sol?
Terra is enough for most work - the gaps are small. Escalate to Sol when a mistake is genuinely costly, on the longest documents, or on multi-step agentic runs where Sol's lead widens (53.6 vs 50.4 on Agents' Last Exam).
Terra or Luna for long documents?
Terra, clearly. It holds 89.6% long-context recall where Luna drops to 41.3%. If your input is long, Terra is the cheapest tier worth trusting.
Is Terra fast enough for interactive use?
Yes - it targets roughly 500-900ms, which is comfortable for conversational assistants. Luna is the one to pick if you need sub-400ms.
Do I need a paid plan?
Yes - Terra is a premium model and needs a Pro plan. Start the 7-day free trial to use it, cancel any time before day eight.
