AI Models / REPORT 001
GPT-5.6 Luna, Terra, and Sol explained
OpenAI splits GPT-5.6 into three tiers, ranging from fast everyday work to complex tasks that demand deeper reasoning.
GPT-5.6 arrives as a family rather than a single model. Luna, Terra, and Sol target different balances of task complexity, speed, and cost.
The idea is straightforward: use a fast compact model for routine work, a balanced model for production, and the highest tier when a problem genuinely needs deeper reasoning.
Luna for fast, repeatable work
Luna is positioned for high-volume everyday tasks such as chat replies, document summaries, classification, and first-draft content.
Its appeal is speed and accessibility, particularly in routine workflows and products serving many users.
Terra for balanced production work
Terra balances capability, accuracy, and cost. It is aimed at organisational work, document reasoning, data analysis, and moderately complex coding.
For many teams, it is likely to be the practical default when a model must be versatile enough for production without paying for the top tier on every request.
Sol for the hardest problems
Sol focuses on deeper reasoning and multi-step problem solving across research, engineering, complex planning, advanced coding, and cybersecurity.
The more important shift is not simply a larger model, but a workflow designed to spend more effort on difficult problems and coordinate agent-like work at larger scale.
Choose for the task, not the badge
Early access was described as limited and centred on the API and Codex before broader availability.
The practical lesson is to map each workflow first: routine volume goes to Luna, balanced production work to Terra, and genuinely difficult multi-step work to Sol.
SOURCES / 01