OpenAI reorganized its flagship around GPT-5.6, splitting it into three durable capability tiers — reportedly named Sol (frontier reasoning), Terra (balanced), and Luna (fast and cheap) — after the public GPT-5.6 launch on July 9, 2026. Alongside it came ChatGPT Work, an agentic system built on GPT-5.6 designed to execute complex, multi-hour projects rather than answer single questions.
The tiering logic
Reported pricing puts Sol around $5/$30, Terra ~$2/$12, and Luna ~$0.20/$1.20 per million tokens — and Luna became the free ChatGPT default with a steep (reported ~80%) price cut. This is the "router" idea baked into the product line: cheap model for the easy majority, frontier model for the hard minority, chosen by tier.
Naming tiers Sol/Terra/Luna instead of confusing version numbers is itself a strategy — it makes "pick the right model for the job" legible to normal users.
ChatGPT Work and long-horizon agents
ChatGPT Work is the consumer-facing face of a trend we keep seeing: agents that run for hours, decomposing and executing multi-step projects. It fits the same lineage as OpenAI's Astra research system and enterprise agent rollouts — the frontier is shifting from single answers to sustained, tool-using work.
Confirmed vs. reported
The GPT-5.6 launch and ChatGPT Work are confirmed product moves; the exact tier names, price points, and the "80% cut" figure come from reporting and may be refined. Long-horizon agent reliability, as always, is best judged in real use, not launch demos.
Why it matters
Two signals here define 2026: aggressive price cuts on capable models (the frontier getting cheaper), and products built around agents that do sustained work. Together they push AI from a chat box toward an execution layer — cheaper to run and able to carry out projects, not just conversations.
