List your company

Speech Reasoning on MMAR-Speech

72.11Speech AccuracyGemini 2.5 Flash
22.439635.334848.2361.1252Mar 19, 2026

Evaluation results

MethodLinks
Gemini 2.5 FlashSize=-2026.0372.11
GPT-4o AudioSize=-2026.0370.41
TGSRSize=7B2026.0364.63
Audio-ThinkerSize=8.4B2026.0364.29
Qwen2.5-omni + Reasoning SFTSize=7B2026.0362.93
Qwen2.5-omni + Only STASize=7B2026.0361.22
Qwen2.5-omniSize=7B2026.0359.86
Audio Flamingo 3Size=7B2026.0357.48
Audio-CoTSize=8.4B2026.0334.01
Audio-ReasonerSize=8.4B2026.0332.99
SALMONNSize=7B2026.0324.35