GPT-5.6 Terra — Specs, Pricing & Limits

GPT-5.6 Terra sits in the middle of its family — priced well under Sol, well above Luna — and it's worth knowing exactly what that middle position does and doesn't buy you, because it isn't a straightforward "less capable, less expensive" tradeoff on every axis.

Context window, maximum output length and knowledge cutoff are all identical to Sol's — Terra doesn't trade any of those away for its lower price. What actually differs is the per-token rate itself, at a meaningful discount to Sol across every service tier and both the short- and long-context bands, and Terra also shares Sol's exact rate-limit group rather than sitting in a cheaper, more restricted one of its own. That combination — same window, same throughput tier, lower price — makes Terra worth checking seriously for any workload currently defaulting to Sol purely because Sol is what the CLI ships with, rather than because a session specifically needs Sol's particular strengths.

The published long-context repricing rule applies to Terra the same way it applies to Sol: crossing the threshold reprices the entire request, not just the excess, so a workload planned against Terra's lower base rate still needs to respect the same cliff Sol's own page describes.

The table below is computed directly from this site's facts module — pricing across every published service tier, the long-context rule, and the published rate-limit table.