OpenAI has released the GPT-5.6 model family, introducing three new sizes—Sol, Terra, and Luna—alongside a new "ultra" effort level that coordinates four agents in parallel by default.
- GPT-5.6 Terra performs just above Fable 5, while Luna outperforms Opus 4.8.
- The new models achieve higher performance at lower cost, using roughly one-third of the time and half as many output tokens compared to competitors.
- GPT-5.6 sets new state-of-the-art results on Terminal-Bench 2.1 and DeepSWE for complex command-line workflows and long-horizon engineering.
The launch coincides with updates to ChatGPT Work and the Codex desktop app, marking a significant expansion of OpenAI's product stack.