ElevenLabs ships Eleven Realtime speech-to-speech with v1 Max benchmark claims
ElevenLabs released two speech-to-speech models called Eleven Realtime for real-time AI conversations. The Decoder reports that the v1 Max variant scores 98.8 percent on the Big Bench Audio benchmark with a 0.57-second response time. ElevenLabs says no model on the Artificial Analysis leaderboard is both faster and more accurate, positioning the lineup for responsive voice AI. Builders gain a new vendor option to compare against existing real-time stacks, but the cited material relies on company-reported benchmark figures rather than independent testing. The excerpt does not spell out pricing, availability, or how the two models differ beyond the Max metrics, and real-world conversation quality may diverge from a single benchmark score under varied network and load conditions.
ElevenLabs ships Eleven Realtime speech-to-speech with v1 Max benchmark claims
Elevenlabs released two new speech-to-speech models called Eleven Realtime. The Max variant scores 98.8 percent on the Big Bench Audio benchmark with a 0.57-second response time.
Key takeaway
A 98.8 percent Big Bench Audio score and sub-second latency claim are evaluation prompts, not proof your voice agent will outperform incumbents in production.
What happened
According to The Decoder, ElevenLabs released two new speech-to-speech models branded Eleven Realtime, aimed at real-time AI conversations where low-latency audio exchange matters.
The Decoder reports that the v1 Max variant scores 98.8 percent on the Big Bench Audio benchmark with a 0.57-second response time, and that ElevenLabs claims no Artificial Analysis leaderboard model is both faster and more accurate.
Evidence
ElevenLabs released two Eleven Realtime speech-to-speech models for real-time AI conversations.
The Decoder · attributed
Elevenlabs released two new speech-to-speech models called Eleven Realtime.
The v1 Max variant is reported at 98.8 percent on Big Bench Audio with 0.57-second response time.
The Decoder · attributed
The Max variant scores 98.8 percent on the Big Bench Audio benchmark with a 0.57-second response time.
ElevenLabs asserts a speed-accuracy lead versus other models on the Artificial Analysis leaderboard.
The Decoder · attributed
Elevenlabs says no model on the Artificial Analysis leaderboard is both faster and more accurate.
Why it matters
Voice-agent builders now have a newly public ElevenLabs real-time speech-to-speech option to bench against rivals, but shipping decisions still depend on workload-specific latency and accuracy outside one vendor benchmark.
Limits and uncertainties
The packet provides only excerpt-level coverage with vendor-reported Big Bench Audio and response-time figures, not independent verification.
Pricing, availability, and differences between the two Eleven Realtime models beyond the Max variant metrics are not described in the cited material.
Practical implications
Run your own latency and accuracy evaluations on representative voice-agent traffic before replacing an existing real-time speech stack with Eleven Realtime.
Treat Artificial Analysis leaderboard comparisons as ElevenLabs positioning until you reproduce results on your infrastructure and audio conditions.
What to watch
Independent benchmarks or customer reports that measure Eleven Realtime latency and quality outside Big Bench Audio and Artificial Analysis leaderboard framing.
