DATASETS2SCOVAL BENCHMARK ROTATIONACTIVE

s2s-workflow-bank-extra-hard-v1 speech-to-speech benchmark dataset

s2s-workflow-bank-extra-hard-v1 is a speech-to-speech test set in Coval's daily rotation — every measured speech-to-speech model runs the same fixed inputs from it.

Models measured
10
last 30 days
Current leader · V2V#1 / 10
897ms
Dialog Rsn 1via PolyAI

How models rank on s2s-workflow-bank-extra-hard-v1

Full S2S dashboard
Voice-to-Voice Latency on s2s-workflow-bank-extra-hard-v1Milliseconds · lower is better · 30-day average on this datasetEvery S2S model measured on the s2s-workflow-bank-extra-hard-v1 dataset, ranked on Voice-to-Voice Latency.
median of all models · 1407 ms
S2S models on the s2s-workflow-bank-extra-hard-v1 dataset over the last 30 days, ranked on Voice-to-Voice Latency.
#ModelHostV2VSamples
1Dialog Rsn 1PolyAI897 ms71
2PrismRime1054 ms27
3Qwen3.8 Omni Flash RealtimeAlibaba1211 ms120
4GPT-Live-1OpenAI1321 ms90
5Phonic 1Phonic1343 ms96
6Eleven Realtime v1 MiniElevenLabs1470 ms120
7Eleven Realtime v1 MaxElevenLabs1472 ms118
8Gemini 3.1 Flash Live (Preview)Google1514 ms143
9Gemini 3.8 LiveGoogle1670 ms144
10Grok Voice 3.0SpaceXAI1824 ms118
Under-sampled models are excluded; tied models share a place.

Metrics reported for this dataset

Same datasets, prompts and metric definitions for every model, measured by Coval’s open-source runner. Full methodology on the overview.

Evaluate your own voice agent

Use Coval to test your production configuration, prompts and calls—not only the public benchmark endpoints.

Book a Demo