Type whats my model _ (or list models _) and the _ answers with the exact model each surface is using right now; the opencues models CLI prints the same effective routing plus the full provider catalog. Both read from dispatch truth - the real per-bucket resolution after every override - so what they show can't drift from what actually runs.
Two ways to ask
| Where | What it shows |
|---|---|
whats my model _ / list models _ (any text field) | the effective model for the surface you're in, answered inline |
opencues models (CLI) | the effective per-bucket routing plus the full catalog of providers and known models |
Why "effective" matters
OpenCues routes in three buckets - cues, blanks, and auditors - each of which can pin its own provider and model, on top of a global default and an auto-fallback order. Reading the config file alone won't tell you what actually runs, because a bucket override, a per-feature override, or a fallback can change it. The model blank and opencues models both run the same resolution walk that dispatch uses, so "what's my model?" and "what did we just call?" are structurally unable to disagree.
Related controls
To change what's routed, see how to switch models; for the list of providers OpenCues supports, see supported models. opencues doctor also surfaces the effective routing per bucket alongside a health check.
Related: How to switch models · Best model for inline AI · Supported models · GitHub repo