List your company

Speech-to-text reasoning and semantic understanding on VoiceBench (test)

4.8Alpaca EvalWhisper + GPT-4o
1.89842.65173.4054.1583Dec 31, 2025

Evaluation results

MethodLinks
Whisper + GPT-4oSystem Paradigm=Cascaded System, ASR Module=Whisper-large-v3, LLM Backbone=GPT-4o2025.124.84.474.6275.7787.298.2776.5192.9781.6987.8
GPT-4oSystem Paradigm=End-to-end Speech-LLM2025.124.784.494.5875.584.198.6576.0289.2380.2586.75
Qwen3-Omni-30BSystem Paradigm=End-to-end Speech-LLM2025.124.744.544.5876.980.499.377.889.768.185.49
Qwen2.5System Paradigm=Text Only Model, Backbone=Qwen2.5-7B-Instruct2025.124.664.554.6262.038099.0470.1484.8471.5782.69
Whisper + Qwen2.5System Paradigm=Cascaded System, ASR Module=Whisper-large-v3, LLM Backbone=Qwen2.5-7B-Instruct2025.124.644.334.2158.552.8598.2763.9978.246976.05
Qwen2.5 (TN)System Paradigm=Text Only Model, Backbone=Qwen2.5-7B-Instruct2025.124.614.534.5663.8456.398.8566.1174.0764.5177.52
AZEROSSystem Paradigm=End-to-end Speech-LLM2025.124.444.183.9160.2256.398.6561.2972.0959.0173.13
GLM-4-VoiceSystem Paradigm=End-to-end Speech-LLM2025.123.973.423.1836.9852.888.0825.9253.4139.7556.48
Qwen2.5-OmniSystem Paradigm=End-to-end Speech-LLM2025.123.883.773.5246.7563.797.3140.1981.5461.4568.26
Phi-4-multimodalSystem Paradigm=End-to-end Speech-LLM2025.123.813.823.5639.7861.810045.3565.9342.1964.32
DeSTA2.5System Paradigm=End-to-end Speech-LLM2025.123.732.523.346.4762.497.6965.4772.7558.5666.04
Qwen2-AudioSystem Paradigm=End-to-end Speech-LLM2025.123.423.292.7631.655399.0426.3548.3536.1453.77
MoshiSystem Paradigm=End-to-end Speech-LLM2025.122.011.61.315.6447.444.2310.1225.9324.0429.51