Benchmark speech-to-speech models
Test and compare speech models on the same phone-call scenarios, including healthcare appointments and Medicare intake.
Cekura Bench
Compare speech-to-speech models on live phone calls
The overview
What is Cekura Bench?
Cekura Bench is a benchmarking feature for comparing speech-to-speech models in live phone-call scenarios. It gives voice AI teams a way to evaluate models under conversational conditions. The feature is part of Cekura, a broader platform for testing and monitoring voice and chat agents across development and production.
Around the internet
What people are saying
Public experiences, questions, and discussions.
No discussions collected yet.
Inside the toolkit
Key features
Compare reliability, speed, and cost
The leaderboard reports repeat-run reliability, success rate, data accuracy, stalled calls, response time, and cost per minute.
Filter and inspect rankings by scenario
Filter the leaderboard by suite or call type and open a model for its detailed results.
Public methodology and benchmark data
The site links test cases, agent definitions, benchmark code, recordings, and downloadable JSON data.
Where it fits
Use cases
- Compare speech-to-speech modelsCompare models on the benchmark’s clinic and Medicare call scenarios using reliability, response time, and published cost.
- Evaluate healthcare phone workflowsReview results for appointment booking, changes, cancellations, and Medicare information collection and routing.
- Review how benchmark results were producedInspect the published method, cases, code, and recordings behind the rankings.
Keep exploring
