Alibaba's Qwen Audio 3.0 TTS Plus tops the competition in the text-to-speech rankings

July 21, 2026 | Source: THE DECODER

Tags: Qwen, Alibaba, text-to-speech, TTS, Qwen-Audio, Speech Arena, multilingual

Alibaba's Qwen-Audio-3.0-TTS-Plus leads Artificial Analysis' Speech Arena with an Elo of 1,236, narrowly edging SpeechifyAI's Simba 3.2 — but its 16 chars/sec generation speed trails Sonic 3.5 (120 chars/sec) by a huge margin, limiting its practical value for latency-sensitive workloads.

Details

Alibaba's new TTS model tops the Artificial Analysis Speech Arena leaderboard for provider voices with an Elo of 1,236, beating SpeechifyAI's Simba 3.2 (1,234) by just two points. Gemini 3.1 Flash TTS (1,214) and ElevenLabs Sonic 3.5 (1,207) follow behind. The model ships in two variants: Flash targets real-time interaction at approximately 300ms latency, while Plus prioritizes output quality. Language coverage spans 16 languages, including underserved markets like Tagalog, Malay, Thai, Vietnamese, and several Chinese dialects. Users can control speaking style through natural language prompts or embedded tags such as [angry] or [giggles], and Alibaba claims improved voice cloning on noisy or echo-heavy reference recordings. The critical limitation is generation speed. At 16 characters per second, Qwen-Audio-3.0-TTS-Plus is nearly 8x slower than Simba 3.2 (30.2 chars/sec) and 7.5x slower than Sonic 3.5 (120 chars/sec). For asynchronous or batch use cases the quality lead may justify the tradeoff; for real-time applications it does not. Pricing is 7.60 per million characters through Alibaba Cloud Model Studio.