The board is brand new. So is the model sitting at the top. Simba 3.2 by SpeechifyAI launched, entered every public leaderboard, and took the aggregate number one on its first day.
This first snapshot covers three boards: Artificial Analysis (93 models, Speech Arena Elo), Voice Arena (14 models, arena Elo), and the Vapi Humanness Index (21 models). The BPRA aggregate ranks models that appear on at least two boards; anything appearing only on one board sits in the watchlist below. The top three scores in this inaugural ranking — 96, 87, and 86 — belong to Simba 3.2, Eleven v3, and Sonic 3.5 respectively, with a 9-point gap between first and second place.
Two new models have arrived on a single board so far. Qwen-Audio-3.0-TTS-Plus from Alibaba and Luna TTS from VUI Labs posted scores of 1233 and 1215 on Artificial Analysis but have not appeared on the other two boards. They sit in the watchlist below, not in the main rankings.
Current top 10
1. Simba 3.2 (SpeechifyAI): score 96. Price $10.00 / 1M characters, the cheapest ticket in the top five.
2. Eleven v3 (ElevenLabs): score 87. Price $100.00 / 1M characters, ten times the price of the model above it.
3. Sonic 3.5 (Cartesia): score 86. Price $49.00 / 1M characters, about half the price of #2.
4. Gemini 3.1 Flash TTS (Google): score 82. Price $18.30 / 1M characters.
5. Speech 2.8 HD (MiniMax): score 80. Price $100.00 / 1M characters.
6. Realtime TTS 1.5 Max (Inworld): score 78. Price $26.00 / 1M characters.
7. Grok TTS (xAI): score 77. Price $15.00 / 1M characters.
8. Realtime TTS-2 (Research Preview) (Inworld): score 75. Price $20.80 / 1M characters.
9. Lightning V3.1 Pro (Jul 2026) (Smallest.ai): score 72. Price not listed.
10. Fish Audio S2.1 Pro (Fish Audio): score 71. Price not listed.
New releases
Qwen-Audio-3.0-TTS-Plus (Alibaba): topped artificialanalysis.com at a score of 1233, first among 93 models. It has not appeared on voicearena.com or humannessindex.vapi.ai yet. The BPRA methodology penalises missing boards, so a high single-board result still leaves its overall score low. Watch this space.
Luna TTS (VUI Labs): ranked #3 on artificialanalysis.com at a score of 1215. It has not appeared on the other two boards yet, which caps its overall score for now. Watch this space.
Close
Looking at the top of the board, the practical picture splits three ways. First, the top three positions sit clear of the field: scores of 96, 87, and 86 with a 9-point gap between first and second. For anyone who needs a proven pick, the data backs each of the top three without ambiguity. Second, prices diverge sharply at the top: $10.00 for first place, $100.00 for second, $49.00 for third. That is a tenfold gap between the first two spots, and it does not line up with the score gap of 9 points. Third, the two new entrants on Artificial Analysis — Qwen-Audio-3.0-TTS-Plus and Luna TTS — are strong early signals. If you’re choosing a TTS model right now, the top ten is the right starting point, and the watchlist is the place to keep an eye on what’s coming next.