September 28, 2026
A new leader in voice generation: Eleven v4 reaches 1,319 Elo
On Sep 28, Eleven v4 reached 1,319 Elo in Provider Voice and took first place in the Artificial Analysis rankings. The model surpassed Sonic 3.6 with 1,276 Elo and Gemini 3.8 Flash TTS with 1,267 Elo across 1,674 appearances. ElevenLabs charges $80 per 1 million characters.
Artificial Analysis
@artificialanlys
ElevenLabs Eleven v4 took first place in the Artificial Analysis Provider Voice TTS Arena rankings and the pronunciation consistency benchmark, while placing second in Controlled Voice. It surpassed Cartesia's Sonic 3.6 and Google's Gemini 3.8 Flash TTS in Provider Voice. Eleven v4 is ElevenLabs' latest speech synthesis model. It supports 90+ languages versus 70+ for Eleven v3. Key points: ➤ Provider Voice: Eleven v4 took first place with an Elo of 1,319 (+19/-19) across 1,674 appearances. Sonic 3.6 scored 1,276, and Gemini 3.8 Flash TTS scored 1,267. Eleven v4 ranked first in all four categories: customer service, assistants, knowledge sharing, and entertainment. ➤ Controlled Voice: for comparison, all models use the same custom voice. Eleven v4 took second place with an Elo of 1,157 (+16/-16) across 1,483 appearances. It trailed Alibaba's Qwen-Audio-3.1-TTS-Plus with 1,178 Elo, but substantially outperformed Eleven v3 with 1,073 Elo. ➤ Pronunciation consistency: Eleven v4 scored 91.7% — the highest result we have measured. Gemini 3.8 Flash TTS scored 89.5%, Gemini 3.1 Flash TTS scored 88.2%, and Eleven v3 scored 85.6%. ➤ Price: Eleven v4 costs $80 per 1 million characters. Sonic 3.6 costs $49, and Gemini 3.8 Flash TTS costs $16.49. ➤ Speed: Eleven v4 generates 73.4 characters per second. Eleven v3 generates 42.5. Details and audio samples below 🧵
· 3.6K views
Previously, Eleven v3 supported 70+ languages. Eleven v4 now supports 90+ languages and rose from 1,073 to 1,157 Elo in Controlled Voice. For pronunciation consistency, v4 scored 91.7% versus 85.6% for the previous version.
In Provider Voice, v4 took first place in all four categories and surpassed Sonic 3.6 and Gemini 3.8 Flash TTS. ElevenLabs charges $80 per 1 million characters, Cartesia charges $49, and Google charges $16.49.
Eleven v4 generates 73.4 characters per second versus 42.5 for Eleven v3.
