New featureModels ·
GPT-5.6 Sol, Terra, and Luna
GPT-5.6 SolGPT-5.6 TerraGPT-5.6 Luna
What it does
OpenAI's GPT-5.6 family arrived in three variants, each selectable per agent.
Why it matters
Three points on the speed-versus-depth curve from one provider, so you can tune cost per agent without changing how your prompts behave.
The specs
OpenAI shipped GPT-5.6 as three separately selectable models rather than one model with a size dial, so you pick the point on the speed-versus-depth curve per agent.
- Shared specs
- All three take 1,050,000 tokens in and produce up to 128,000 out, read text and images, and support Actions.
- Sol and Terra
- Carry the 'smartest' tag — the two to reach for on hard reasoning.
- Luna
- The lighter sibling for everyday work.
- Reasoning
- Fast ($$), Smart ($$$), or Smartest ($$$$) on all three.
- Plans
- Gold and Pro.
Because the three share a context window and prompt behaviour, moving an agent between them is a one-dropdown change — you can tune cost after launch without re-testing your prompt.