Model management
Changing models
Choose Configure all agent models, Configure individual agents, or Manage presets in the menu. Each flow offers a searchable model list sourced from opencode models, then shows a preview before writing.
For scripting:
omo-model-slim set openai/gpt-6.1-sol --all --dry-run
omo-model-slim set openai/gpt-6-luna --agent presets/openai/oracle --yes
Targets accept the exact model path shown by omo-model-slim current, the agent name, or a slash-separated path such as presets/openai/oracle.
Fallback preservation
Only the primary model is replaced. Every element of an ordered fallback chain is left alone:
"oracle": {
"model": ["openai/gpt-6.1-sol", "openai/gpt-6-luna"],
"variant": "high"
}
Changing the primary with --agent presets/openai/oracle produces ["new/model", "openai/gpt-6-luna"]. If a chain uses object entries with inline variants, only the id is rewritten so the inline variant survives.
Verification
omo-model-slim verify
Each configured model ID is checked against the live OpenCode registry. Verification is a registry check only: it does not authenticate providers or make requests.
Model discovery and catalog refresh
Refresh model catalog re-runs opencode models and opencode auth list. Free-labeled models appear first, then authenticated providers, then everything else. If discovery fails, the error is shown in the dashboard instead of crashing the app.