Posts tagged with "Benchmarks"
4 posts found

NVIDIA Nemotron 3.5 Transcribe Is #1 Overall on Our Speech-to-Text Benchmark
NVIDIA Nemotron 3.5 Transcribe has the lowest overall word error rate on Cekura Bench: 2.28% across public clips and private phone calls, first of 21 speech-to-text models.

Sidhant Kabra
Wed Nov 04 2026

Cekura Bench vs. τ-bench: Benchmarking Voice Agents the Way They Actually Run
τ-bench tests whether a model can do the task. Cekura Bench tests your deployed voice agent on real audio calls, three runs each. See how the two compare.

Sidhant Kabra
Thu Oct 15 2026

Gemini Live API 2026: Models, Benchmarks & Gemini 4
Gemini Live API guide: the newest models, current pricing, whether Gemini 4 works with it, and how each model scores in Cekura's call benchmark.

Janhvi Nandwani
Tue Nov 03 2026

Tau-Bench Explained: Tau2, Tau3, and Pass^k (2026)
Tau-bench explained: how Sierra's tau benchmark scores AI agents, what tau2-bench and tau3 bench changed, and why pass^k exposes inconsistent agents.

Janhvi Nandwani
Tue Nov 03 2026