GPT-5.6 Is Live — Luna, Terra and Sol Scale to the Job

July 10, 2026

You can now choose the GPT-5.6 model that matches the work: Luna for fast, high-volume tasks, Terra for everyday coding and agent runs, or Sol for the hardest reasoning and multi-step workflows. All three share a 1.05M-token context window, support tools and Swarm, and automatically spend more reasoning effort only when the task calls for it.

What you can do

• Pick GPT-5.6 Luna for speed and volume — chat, classification, extraction, lightweight research, and worker tasks at the Workhorse 3× band.
• Pick GPT-5.6 Terra for balanced daily work — coding, reasoning, and agentic workflows at the Premium 6× band.
• Pick GPT-5.6 Sol for maximum depth — complex analysis, difficult coding, and long multi-step agent runs at the SOTA 12× band.
• Keep large projects in one context — each model accepts up to 1.05M tokens of context and can produce up to 128K output tokens.
• Let effort follow complexity — Alfrada scales reasoning and answer detail from low, through medium, to high. The hardest tasks can activate GPT-5.6's pro reasoning mode automatically.
• Use the family throughout Alfrada — Luna, Terra, and Sol all support tool calling, Auto routing, and Swarm. Luna can also serve as a cost-efficient worker inside multi-agent runs.

Where this shows up

• You have hundreds of documents to classify or extract. Choose Luna to keep throughput high without giving up long context or tools.
• You are refactoring a codebase or running an everyday research workflow. Choose Terra for the middle ground between speed, cost, and reasoning depth.
• You have a difficult strategic decision, an ambiguous engineering problem, or a long agent plan that must hold together across many steps. Choose Sol for the strongest GPT-5.6 configuration.

Try it

• "Use GPT-5.6 Luna to classify these files by topic, urgency, and owner, then return a concise action list."
• "Use GPT-5.6 Terra to inspect this repository, plan the refactor, implement it, and verify the risky paths."
• "Use GPT-5.6 Sol to challenge this market-entry plan, research the missing evidence, and produce a board-ready recommendation."

Heads up

• The three models use different token bands. Luna is Workhorse 3×, Terra is Premium 6×, and Sol is SOTA 12× against your plan quota.
• Auto can select all three. Luna and Terra sit in Auto's Workhorse tier, while Sol sits in SOTA. Pin a model when you want a specific cost and capability profile.