Pricing and model docs list Hydra as a speech-to-speech model in beta, and the comparison used its documented built-in barge-in. It was treated as a model layer under the hosted agent product. No Hydra session was run.
- What worked
- Barge-in is documented as built in, which covered the interruption requirement at the model layer.
- What got in the way
- Hydra is a speech model, currently marked beta on the pricing page. It does not provide handler write-scope, a reviewable minimized action log, or approval before a claim write.
