Cekura Bench
Updated
About Cekura Bench
Cekura Bench tests and ranks real-time voice AI models on how well they perform as phone agents.
Developers and teams building voice AI applications use Cekura Bench to compare how different speech-to-speech models perform on actual phone tasks. It helps you pick the most reliable, fastest, or most cost-effective model for your needs.
- Tests 9 real-time voice models including GPT, Gemini, and Grok
- Ranks models on reliability, accuracy, response time, and cost
- Public transcripts and scores for every test call
- Voice agent and STT benchmarks; TTS coming soon
Good to know: All test calls, transcripts, tool calls and scores are public; agent and test cases are available on GitHub.
Key features
- Comprehensive phone testing: Tests models across 82 real-world scenarios (clinic booking and Medicare intake) with three runs each.
- Transparent rankings: Models ranked on reliability, data accuracy, stalled calls, response time, and cost with full call data visible.
- Multiple benchmark types: Covers speech-to-speech, voice agent, and STT benchmarks with TTS benchmarks planned.
At a glance
- Category: AI Coding
- Type: Text-to-speech & voice
- Jobs: Produce Audio & Voice
- Evidence: New launch on Product Hunt (a few reviews)
- AIME Score: 55 of 100
- Launched: Oct 8, 2026 on Product Hunt · 100+ upvotes
- Website: cekura.ai
Alternatives to Cekura Bench
- Krisp (AIME Score 89)
- ElevenLabs (AIME Score 82)
- Fliki (AIME Score 76)
- WellSaid Studio (AIME Score 72)
- LOVO AI (AIME Score 71)
- Fish Audio (AIME Score 70)
More Text-to-speech & voice tools
Sources: Found on Product Hunt (launched Oct 2026) · Details checked on the tool’s own website · Researched by AI Made Easy
This is a tool profile with the essentials. Spotlights, our in-depth reviews with screenshots and video, are added every day, Proven tools first. How we pick Spotlights
Is this your tool? Claim, update or remove this listing, or get a badge