Speechify's SIMBA 3.2 Ranks First on Artificial Analysis Text-to-Speech Leaderboard

Speechify's streaming text-to-speech model SIMBA 3.2 ranks first on the Artificial Analysis text-to-speech leaderboard as of July 2026, ahead of models from ElevenLabs, Cartesia, OpenAI, and Google DeepMind. On Voice Arena, a blind-listener benchmark, SIMBA 3.2 is in a statistical tie for second on the U.S. English leaderboard.

Neither leaderboard is operated by Speechify, and neither relies on self-reported scores. The Artificial Analysis leaderboard applies a consistent objective evaluation across commercially available models. Voice Arena uses native-speaker panels who hear two audio clips from identical text and vote for the more natural clip.

Arena aggregates votes into Elo ratings and covers six languages. It uses a balanced voice slate per model and evaluates text written for production contexts. The methodology was developed with input from Prof. Shinji Watanabe of Carnegie Mellon University.

The company lists SIMBA 3.2 at $10 per one million characters on its entry tier and $6 per one million characters on its Scale tier. A company statement said the model resulted from treating cost, quality, and latency as a single problem from the start, driven by a consumer platform that serves more than 60 million users.

The same model is available through a REST API and first-party TypeScript and Python SDKs. It supports streaming with lower time-to-first-byte than previous generations, fine-grained emotional control, SSML prosody, instant voice cloning, and native-quality speech across more than thirty locales.

Additional languages and a lower-cost version are on the roadmap.