Topping the Charts in 20+ Languages, VoxCPM 2 Refreshes the MiniMax Leaderboard

According to scoring on public datasets, VoxCPM 2 — the new-generation end-to-end TTS speech model that ModelBest released and open-sourced on April 6 — stands out on MiniMax-Eval multilingual similarity.

It holds first place across 20 languages and continues to lead on Cantonese, ahead of Qwen3-TTS, Eleven Labs, and MiniMax.

Additional Images