SOTA Speech Understanding on MMSU and PapersWithCode

81.3Accuracy

Gemini-3.1-Pro

Updated 16d ago

Evaluation Results

Method	Links
Gemini-3.1-Pro 2026.06		81.3
Gemini-3-Pro 2026.06		81.28
MOSS-Audio-8B-Thinking 2026.06		75.52
Step-Audio-R1 2026.06		75.18
MOSS-Audio-4B-Thinking 2026.06		71.2
Qwen3-Omni-30B-A3B-Instruct 2026.06		69
MOSS-Audio-8B-Instruct 2026.06		66.36
Step-Audio-R1.1 2026.06		64.18
SpectCount (Audio Flamingo 3) 2026.06		63.18
Kimi-Audio 2026.07		62.2
MiMo-Audio-7B 2026.06		61.94
Audio Flamingo 3 2026.06		61.92
Audio Flamingo 3 2026.06		61.4
Qwen2.5-Omni-3B 2026.06		61.4
Qwen2.5-Omni 2026.06		61.32
MiniCPM-o-4.5 2026.06		60.96
Audio Flamingo 3 2026.06		60.04
MOSS-Audio-4B-Instruct 2026.06		59.68
GPT4o-Audio 2026.06		58.76
Audio Flamingo Next 2026.06		57.2
DuplexPO 2026.07		56.2
SFT Baseline 2026.07		54.9
Kimi-Audio 2026.06		54.74
SpectCount (Qwen2-Audio-Instruct) 2026.06		54.24
Qwen2-Audio-Instruct 2026.06		53.27
Baichuan-Audio 2026.07		53.2
Gemma-4-E2B-it 2026.06		52.8
AVLM-2B 2026.06		51.9
Qwen2-Audio-Instruct 2026.06		48.44
GLM-4-Voice 2026.07		39.8
Qwen2-Audio 2026.07		35.7
SALMONN-omni 2026.07		30
Freeze-Omni 2026.07		28.1
SALM-Duplex 2026.07		26.3
Moshi 2026.07		24