Voice AI benchmarks
Independent benchmarks for voice platforms, speech to speech and speech to text models, scored on live calls and real call audio.
Popular comparisons
To run these tests on your own agent, see our guide to AI voice agent testing tools.
- Universal 3.6 ProvsUniversal 3.5 Pro
- Universal 3.6 ProvsScribe Pro
- Universal 3.6 ProvsScribe
- Grok Voice 3.0vsGPT Realtime 2.1
- Grok Voice 3.0vsGPT-Live 1 + GPT-6 Sol
- Grok Voice 3.0vsAzure Realtime
Every pair: speech to text, speech to speech, agent platforms.
Voice AI benchmark FAQ
What does Cekura Bench compare?
Cekura Bench compares voice agent platforms, realtime speech-to-speech models, production conversation quality, and speech-to-text models under documented test conditions.
Cekura Bench compares voice agent platforms, realtime speech-to-speech models, production conversation quality, and speech-to-text models under documented test conditions.
The benchmarks measure reliability, task completion, data accuracy, response time, voice quality, interruption handling, and speech recognition accuracy where each metric applies.
Yes. Cekura Bench includes dedicated head-to-head comparison pages for speech-to-speech and speech-to-text models tested under the same conditions.
Pass³ is the share of scenarios where all three retained test calls passed. A scenario must succeed repeatedly rather than only once.
Yes. Each benchmark explains its methodology and links to supporting reports, test cases, or source material where available.