Fish Audio has upgraded the weights of Drama 3, its open-weight speech model, and the independent rankings moved with it. In a post on X on 29 October 2026, the company said Drama 3 is now the top open-weight model in both speech arenas run by Artificial Analysis, the firm that benchmarks AI models.
In the Provider Voice arena, Drama 3 ranks sixth among all models, open or closed, and gained 38 Elo points over the preview release. In the Controlled Voice arena it leads the open-weight field, up 22 points on the preview. The post also relays Artificial Analysis’s finding that Drama 3 ranks first of all models on its Pronunciation Robustness benchmark.
The post does not say where Drama 3 lands overall in Controlled Voice, or how many votes sit behind the new scores.
A sixth-place finish among every model is a thin gap to paid voice services. Anyone paying per character for text-to-speech now has a reason to run the new weights on their own scripts and compare.
Reported by Fish Audio in a post on X on 29 October 2026.
