ElevenLabs launches Realtime v1 Max: 98.8% audio accuracy with 0.57s latency
Kangwook_Lee · x · 2026-10-21
ElevenLabs released Eleven Realtime v1 Max, an end-to-end realtime voice model scoring 98.8% on Big Bench Audio with a 0.57s time-to-first-audio; per the company, no model on the Artificial Analysis leaderboard is both faster and more accurate.
Prof. Kangwook Lee congratulated the team and noted his lab's Raon SpeechChat still sits on the accuracy-latency Pareto frontier thanks to the lowest latency (100ms, though 58% accuracy), teasing more speech models to come.
More from Multimodal
- Stunning gothic AI-generated image: skull throne, black panthers, ornate detail — tetsuoai · 2026-10-21
- Virtual British rocker built with Suno duets and Dreamina music videos — AIandDesign · 2026-10-21
- Cinematic 'What If You Stole $10M' Animation Made with Higgsfield Supercomputer — EXM7777 · 2026-10-21
- Seedance 2.5 continuity prompt keeps one supercar consistent across five wild AI video shots — Aiden_Tech_Ai · 2026-10-21
- Veo-generated cat CEO ringing the bell for a 'lunch meeting' is the AI meme of the day — Much-Duck1009 · 2026-10-21
- Creator uses refmods with reference images for consistent AI character scenes — Thorozar · 2026-10-21