Gradium models, access and benchmark guide
Gradium model and access guide with a practical US-English evaluation brief and links to publisher materials.

👁 111 reads
A small open English checkpoint can suit a local installation, while a managed speech API may offer a different model and set of controls. Comparing their prices requires timing local inference and identifying the actual hosted version.
Choosing the first job
Use an original museum label with a Roman numeral and a street address. Record pronunciation and local memory usage separately.
Models in the November 17 snapshot
| Model and track | Placement | Reported score |
|---|---|---|
| Gradium TTS (Sep 2026) (AA provider voice) | 28 | 1173 Elo |
| Gradium TTS (Sep 2026) (AA controlled voice) | 24 | 1055 Elo |
| Phonon 31M (English) (AA provider voice) | 44 | 1109 Elo |
Release record. Publisher documentation identifies the product family. The ranked configurations include Gradium TTS (Sep 2026), Phonon 31M (English).
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →Sources
- https://gradium.ai/ (accessed November 17, 2026)
- https://huggingface.co/gradium (accessed November 17, 2026)
- https://artificialanalysis.ai/text-to-speech/leaderboard/provider-voice (accessed November 17, 2026)
- https://artificialanalysis.ai/text-to-speech/leaderboard/controlled-voice (accessed November 17, 2026)
More on this from QUANTSLANT
benchmarks·Fish Audio models, access and benchmark guideFish Audio model and access guide with a practical US-English evaluation brief and links to publisher materials.