# Choosing the right model



> Choose a configured model using the capabilities exposed by your current runtime.



Model names and capabilities are provider- and organization-dependent. Choose from the models exposed by the current MobDial runtime and validate quality, latency, language coverage, limits, and cost against that contract.

## Speech synthesis

- Use only model identifiers returned or documented by the active voice provider/runtime.
- Benchmark representative scripts before selecting a model for production traffic.
- Treat language coverage, input limits, latency, and pricing as runtime/provider properties rather than MobDial-wide constants.

## Speech recognition

- Use the speech-recognition models and options exposed by the configured provider/runtime.
- Verify streaming support, diarization, timestamps, language coverage, and latency in the current contract before depending on them.

Measure end-to-end latency in your own environment; provider inference is only one part of the total path.



---

Powered by MobDial Agents (MobAgents).