GPT-5.6 family explained: Sol vs Terra vs Luna in 2026
OpenAI's GPT-5.6 family has three tiers — Sol, Terra and Luna. Here's how they compare on price, intelligence and what each is actually good at in 2026.
GPT-5.6 family explained: Sol vs Terra vs Luna
OpenAI sells GPT-5.6 in three tiers. They share a name, but they’re very different models with different prices, speeds and strengths. Here’s how to tell them apart — and which one you should actually use.
The three models at a glance
| Model | Price in / out per 1M | Intelligence index | Role |
|---|---|---|---|
| GPT-5.6 Sol | $4 / $20 | 61 | Flagship |
| GPT-5.6 Terra | $2 / $12 | 57 | Balanced |
| GPT-5.6 Luna | $0.20 / $1.20 | 52 | Fast & budget |
All three share a 1M-token context window and were released 9 Jul 2026.
GPT-5.6 Sol — the flagship
GPT-5.6 Sol is OpenAI’s top model and the one practitioners rate highest of the three for coding. It’s the strongest back-end and bug-fixing model in the family (A tier for one-shot work), and it pairs well with sub-agents for ambitious tasks.
The catch: it’s token-hungry. Sol burns through usage limits fast — heavy users report exhausting a weekly subscription cap in about a day — and it needs close guidance for front-end design. At $4 / $20 per 1M tokens it’s the priciest of the three on output (a promotional rate runs through at least November 21, 2026).
GPT-5.6 Terra — the balanced pick
GPT-5.6 Terra is the middle child and the best ChatGPT value for most work: an index of 42 at $2 / $12. It covers the bulk of production tasks without Sol’s token appetite. Practitioners put it in B tier for speed — faster than you’d expect for its class.
GPT-5.6 Luna — the budget tier
GPT-5.6 Luna is OpenAI’s fast, cheap tier at just $0.20 / $1.20 per 1M tokens. It’s ideal for high-volume, latency-sensitive work like classification and extraction. Its weaknesses are real, though: F-tier front-end design and mid-tier one-shot coding make it a poor pick for anything that needs polish.
Which should you pick?
- Front-end design: none of them are great — use a Claude model instead.
- Back-end / bug-fixing: GPT-5.6 Sol.
- Most production work at a sane price: GPT-5.6 Terra.
- Bulk, low-latency pipelines: GPT-5.6 Luna.
Run the numbers yourself in the AI intelligence calculator or see how OpenAI stacks up against the field in ChatGPT vs Claude.