GPT-5.6 Terra
Current version: 5.6 updated
Balanced GPT-5.6 model: code, reasoning, agentic tasks. Half the price of Sol.
What's new in GPT-5.6 Terra 5.6
- 1.05M token context
- Balanced tier between Sol and Luna, roughly half the price of Sol
- Strong in coding, reasoning and agentic tasks
- OpenRouter top categories: Academia 20, SEO 22, Legal 36
- 2.5 USD input, 15 USD output per million tokens
What it's the best tool for
- 77.4 on the Coding Agent Index — above GPT-5.5
- 1.05M-token context with up to 128K output
- 87.4% on Terminal-Bench 2.1 agentic tasks
- Half the price of the flagship Sol tier
- 90% cheaper input on cached prompts
When to reach for something else
- FrontierMath Tier 4 below GPT-5.5: 68.3% vs 72.5%
- More eager than GPT-5.5 to act beyond instructions
- Knowledge cutoff February 2026, no built-in browsing
- OpenAI content filters on sensitive topics
How GPT-5.6 Terra responds
GPT-5.6 Terra
Four scenarios where it pays for itself
More about GPT-5.6 Terra
GPT-5.6 Terra Online — the Balanced Tier of OpenAI's Lineup
GPT-5.6 Terra is the middle tier of the GPT-5.6 family OpenAI released on July 9, 2026 — flagship Sol above it, budget Luna below. Terra is built for everyday production work: coding, reasoning and agentic tasks at roughly half the flagship's price. Try it on NetRoom right in your browser, with no VPN and no separate API keys.
What Terra can do
Independent Artificial Analysis testing puts Terra at 77.4 on the Coding Agent Index — above GPT-5.5's 76.4 and on par with the strongest competing models. It scores 87.4% on Terminal-Bench 2.1 and 63.4% on SWE-Bench Pro, which in practice means it reliably reviews pull requests, fixes bugs across large repositories and carries long multi-step tool chains to completion without babysitting. You get a 1.05M-token context window, up to 128K output tokens and a February 2026 knowledge cutoff.
Terra vs Sol vs Luna
Sol is the frontier tier for the hardest reasoning and math; Luna is the fast, low-cost option for high-volume workloads. Terra sits in between: near-flagship quality on code and agents at half of Sol's price. If you have been running GPT-5.5, Terra matches or beats it on most coding benchmarks while costing noticeably less per request.
Where it already ranks
On OpenRouter, GPT-5.6 Terra holds top-chart positions in Academia (#20), SEO (#22), Legal (#36), Marketing (#40) and Finance (#50) — teams pick it for research and study, SEO copy, legal drafts and financial analysis.
When to pick Terra
Daily coding and code review, autonomous agents, long-document analysis, SEO and marketing content, contract drafts. Go up to Sol for frontier-grade math and the deepest reasoning, or down to Luna for cheap bulk requests. One honest caveat from OpenAI's own system card: Terra is more eager than GPT-5.5 to do more than instructed, so keep agent scopes tight.
How to start
Sign up on NetRoom, top up your balance, pick GPT-5.6 Terra from the model list and start prompting in the browser. Pay as you go — no OpenAI subscription, no key management, just tokens you actually use.
What changed GPT-5.6 Terra
- + Added text model GPT-5.6 Terra (OpenAI).
Version history of GPT-5.6 Terra
| Version | Date | What changed |
|---|---|---|
| 5.6 current |
|
Try GPT-5.6 Terra
right now
Free access to basic models. No card, no obligations.