OpenAI · Model comparison
GPT-5.6: Sol vs Terra vs Luna
The three GPT-5.6 tiers are more alike than most comparisons let on. Same launch day, same 1M context window, same February 2026 knowledge cutoff. What separates them is capability and a 5× price spread.
Side by side
| Tier | Position | Input / 1M | Output / 1M | Context | Released |
|---|---|---|---|---|---|
| Sol | Flagship | $5 | $30 | 1M | 2026-07-09 |
| Terra | Balanced | $2.50 | $15 | 1M | 2026-07-09 |
| Luna | Cost-optimized | $1 | $6 | 1M | 2026-07-09 |
What each tier is for
Sol
$5/$30Flagship
Reasoning-intensive work, software development, multi-step problem solving.
Terra
$2.50/$15Balanced
Everyday coding, reasoning and agentic apps — vendor-positioned between Sol and Luna.
Luna
$1/$6Cost-optimized
Speed-sensitive, high-volume work such as chat and classification.
These positioning notes are how OpenRouter's model pages describe each tier's intended use. They are not benchmark results — no independent head-to-head benchmark of the three tiers is cited on this page, and we will not invent one.
How to pick
Because the context window does not change across tiers, prompt size is irrelevant to the choice — a 400k-token prompt runs on Luna just as it runs on Sol. The decision is only about whether the task needs flagship reasoning. The cheapest useful default is to start every workload on Luna, measure quality on your own evals, and promote only the workloads that fail. Starting at Sol and trying to walk costs down later is the more expensive direction to travel.
Sources
- OpenRouter — GPT-5.6 Sol — retrieved 2026-07-17
- OpenRouter — GPT-5.6 Terra — retrieved 2026-07-17
- OpenRouter — GPT-5.6 Luna — retrieved 2026-07-17
Figures on this page last checked against these sources on 2026-07-17. Vendors change pricing and specs without notice — if a number here disagrees with the vendor's own page, trust the vendor.