GPT-5.6: How to Choose Between Sol, Terra, and Luna in ChatGPT

You open ChatGPT and something feels different. Responses come back differently, sometimes faster, sometimes more thorough. The reason: since July 9, 2026, ChatGPT runs on GPT-5.6, and for the first time, there are three models instead of one.

What is GPT-5.6?

OpenAI released GPT-5.6 not as a single model but as a family of three tiers: Sol, Terra, and Luna. Each tier is optimized for different tasks. The principle: you do not always need the most powerful model. Sometimes a faster, cheaper one will do.

The three models at a glance:

Sol is the flagship. It thinks the deepest but takes longer. Sol is the right choice for:
- Complex analyses and research
- Difficult programming tasks
- Tasks where accuracy matters more than speed
- Multi-step problem-solving that requires logical reasoning

Terra is the all-rounder. It delivers strong results for most everyday tasks and is significantly faster than Sol. Terra works for:
- Writing and revising emails
- Summarizing texts
- Brainstorming and ideation
- General questions and explanations

Luna is the fastest and most efficient model. It responds almost instantly and is suited for:
- Short, simple questions
- Quick translations
- Spelling corrections
- Tasks you repeat frequently

What do you get on each plan?

| Plan | Model in ChatGPT | Reasoning Modes |
|---|---|---|
| Free / Go | GPT-5.5 Instant (default) | No Sol access |
| Plus (20 USD/month) | Sol at Medium and High reasoning | Medium, High |
| Pro (100/200 USD/month) | Sol at all levels + Sol Pro | Medium, High, Extra High, Sol Pro |

Terra and Luna are currently not directly selectable in regular chat. They are available in ChatGPT Work (the business variant) and Codex (OpenAI's coding agent).

How to use reasoning modes in ChatGPT:

If you are a Plus or Pro user, you will find a reasoning mode selector in the chat window. This controls how thoroughly Sol thinks:

- Medium: Sol thinks briefly and responds quickly. Good for most everyday tasks where you still want better quality than the default mode.
- High: Sol takes more time and delivers more thoughtful answers. Ideal for longer texts, analyses, and tasks where details matter.
- Extra High (Pro only): Sol thinks especially long and thoroughly. For the hardest tasks: mathematical proofs, complex code architectures, deep research.

Practical tip: Start with Medium. If the answer is not thorough enough, repeat the question with High. This saves your usage quota, as higher reasoning modes consume more of it.

What has concretely improved?

GPT-5.6 Sol is stronger than GPT-5.5 especially in these areas:

- Coding: On the Terminal-Bench 2.1 benchmark (real-world coding tasks), Sol scores 88.8%, and 91.9% in Ultra Mode. These are the highest scores among all current models.
- Long contexts: All three models process up to 1 million tokens. You can analyze entire books, long documents, or extensive codebases in one go.
- Agent capabilities: Sol can now independently coordinate multiple tools in parallel (Ultra Mode) and even write JavaScript to orchestrate tools. This sounds technical, but it means ChatGPT can handle more complex, multi-step tasks autonomously.

All models compared directly:

| | Sol | Terra | Luna |
|---|---|---|---|
| Strength | Deep reasoning | Balanced performance | Speed |
| Speed | Slower | Medium | Fast |
| API price (input/output) | 5/30 USD per 1M tokens | 2.50/15 USD | 1/6 USD |
| Ideal for | Hard tasks | Everyday work | Quick, simple jobs |
| Context | 1M tokens | 1M tokens | 1M tokens |

Assessment:

GPT-5.6 is not a revolutionary leap over GPT-5.5, but a smart move. Instead of offering a single model for everything, OpenAI gives you a choice: the strongest model for difficult tasks, a fast one for simple ones. For ChatGPT Plus users, the biggest change is in daily use: Sol delivers noticeably better results on complex questions than the previous default. For Free and Go users, everything stays the same for now, though GPT-5.5 Instant remains a strong model.

The most important practical tip: use reasoning modes deliberately. Not every question needs High. Those who consciously switch between Medium and High get better results while using less of their monthly quota.

Sources: openai.com/index/gpt-5-6, help.openai.com/en/articles/20001354, simonwillison.net/2026/Jul/9/gpt-5-6