GPT-6.1 Terra and GPT-6 Luna are OpenAI's efficient flagship tiers. Terra is described as fast and efficient with visual reasoning, precise computer use and natural writing; Luna as OpenAI's most efficient model for focused, high-volume tasks.
Descriptions, limits, prices and cutoffs were re-read from OpenAI's Models page on 2026-11-02.
Key differences at a glance
- Price: $1 / $5 per million input/output tokens for Terra against $0.10 / $0.50 for Luna.
- Same limits: 1.05M-token context and 128K-token output.
- Knowledge cutoff: June 1, 2026 for Terra against May 18, 2026 for Luna.
- Terra's description highlights visual reasoning and computer use; Luna's highlights efficiency for focused, high-volume work.
- Reasoning effort was listed from none through max for both when compiled.
Specification comparison
| Feature | GPT-6.1 Terra | GPT-6 Luna |
|---|---|---|
| Description | Fast, efficient; visual reasoning, computer use, natural writing | Most efficient; focused, high-volume tasks |
| Context / max output | 1.05M / 128K | 1.05M / 128K |
| Price per MTok (in / out) | $1 / $5 | $0.10 / $0.50 |
| Knowledge cutoff | Jun 1, 2026 | May 18, 2026 |
| Reasoning effort | none through max (adds none) | none through max (adds none) |
Shaded cells mean the cited sources did not state the value. We do not fill gaps from memory.
Different jobs
OpenAI's descriptions separate these clearly. Terra is pitched at interactive and agentic work involving images and computer use, plus clear writing. Luna is pitched at narrow, repetitive tasks at scale, such as classification, extraction and short transformations.
Cost at scale
Luna is one-tenth of Terra's price on both input and output. For millions of short calls, Luna's cost advantage is decisive if quality is acceptable. For multi-step agents that operate a computer or reason about screenshots, Terra's capabilities may reduce failed steps, which also costs money.
Limits and freshness
Both list a 1.05M-token context and 128K tokens of output. Terra's knowledge cutoff (June 1, 2026) is slightly newer than Luna's (May 18, 2026), a small difference for most uses.
Choosing
Prototype with Terra where vision or computer use matters; benchmark Luna wherever the task is narrow. Many systems use both: Luna for routing and bulk processing, Terra for the steps that need richer capabilities.
Strengths and limitations
GPT-6.1 Terra
Strengths
- Wider described capability.
Limitations
- Ten times Luna's listed price.
GPT-6 Luna
Strengths
- One-tenth the listed price.
Limitations
- Described for focused, high-volume tasks.
Verified pricing
Terra: $1 / $5. Luna: $0.10 / $0.50. Prices are per million tokens (MTok) as listed on the cited page on the verification date; they can change.
Which should you choose?
- General efficient use: GPT-6.1 Terra — visual reasoning and computer use described
- High-volume focused tasks: GPT-6 Luna — $0.10 / $0.50
Our verdict
Choose Luna when cost per call dominates and tasks are narrow; Terra when you need visual reasoning or computer use.
Frequently asked questions
What is GPT-6 Luna best for?
OpenAI describes it as its most efficient model for focused, high-volume tasks.
Does Terra support computer use?
OpenAI's description of GPT-6.1 Terra mentions precise computer use.
Which is cheaper?
Luna, at $0.10 / $0.50 per million input/output tokens against $1 / $5 for Terra, as listed on 2026-11-02.
Related comparisons
AI Tools & Models
GPT-6 Astra vs GPT-6.1 Sol: Specifications Compared
GPT-6 Astra versus GPT-6.1 Sol
OpenAI's two top models compared on price, context and reasoning efforts per OpenAI's models listing.
Verdict: Sol first; step up to Astra only if tests justify five times the price.
Read comparisonAI Tools & Models
Claude Haiku 5.5 vs GPT-6 Luna: Specifications Compared
Claude Haiku 5.5 versus GPT-6 Luna
Budget-tier API models compared on price, latency label and focus.
Verdict: Near-equal list prices; test accuracy and speed on your task.
Read comparisonMore AI Tools & Models comparisons · All comparisons
How this was checked
Specifications compiled on 2026-11-02 from the sources below. Part of the cited evidence was re-checked on 2026-11-02; see the notes below.
OpenAI Models page re-read on 2026-11-02: descriptions, limits, prices and cutoffs matched. Reasoning-effort rows were not visible in the re-read.
Disclosure: AYORAHUB did not hands-on test these products or services. Nothing on this page is a lab measurement, benchmark or review score; analysis explains what the cited specifications mean in practice.
- OpenAI figures come from the Models listing page; individual model pages may add detail.
Sources
- Models — OpenAI API Docs (manufacturer/official)