Command reference Β· Antigravity CLI
/model
Picks the reasoning model, persisting across sessions.
/model (Configurations category) opens a picker for choosing the reasoning model; according to the CLI reference, the choice applies across sessions until changed again. Since version 1.1.5, the picker is grouped by base model and allows choosing the reasoning effort level via a timeline slider, including an effort indicator in the status bar; the corresponding stable model slugs can also be passed directly to the --model startup flag or to --effort.
According to the official model overview, the Gemini models are available on all plans including Enterprise, while Claude Sonnet 4.6 (thinking), Claude Opus 4.6 (thinking), and GPT-OSS-120b are available on Free, Plus, Pro, and Ultra accounts, but not on Enterprise plans. Which base models are actually available can be checked beforehand with agy models. The model choice is "sticky": a change during a running agent turn only takes effect from the next turn, not immediately.
π THE SAME TASK (βSWITCH MODELβ) ON OTHER PLATFORMS
β WHEN TO USE IT?
Choose a more powerful reasoning model for a more demanding task
Open /model and switch to, e.g., Claude Opus 4.6 (thinking)
Adjust a model's reasoning depth
Open /model, choose a base model, and change the effort level via the slider
β WHEN NOT TO?
Setting the model only for the next startup, not permanently
/model changes the choice persistently across sessions.
Better: Use agy --model <name> at startup
Switching in the middle of a running agent turn and expecting an immediate effect
The choice is sticky β the running turn still completes with the previous model.
Better: Wait until the end of the current turn, or cancel the running execution
SOURCES
- Antigravity CLI docs: command reference β antigravity.google
- Antigravity docs: models β antigravity.google
MORE COMMANDS Β· Antigravity CLI