List your company

Audio Reasoning on MMAR

83.7Average AccuracyGemini-3.1-Pro
16.41233.88151.3568.819Oct 3, 2025Nov 14, 2025Dec 26, 2025Feb 6, 2026Mar 20, 2026May 1, 2026Jun 13, 2026

Evaluation results

MethodLinks
Gemini-3.1-ProSize=–, Model Category=Closed-source models2026.0683.7———————
Gemini-3-ProSize=–, Model Category=Closed-source models2026.0681.73———————
VISASize=Agent2026.0677.471.562.68463.686.281.775
AudioToolAgentSource=Closed Source2025.107773.3363.5982.9963.6483.0381.7179.17
Gemini 3 ProSource=Closed Source2025.1076.472.7361.6584.3572.7381.1978.0583.33
Step Audio R1Size=32B2026.0671.561.252.481.354.680.381.779.2
OmniJigsawvariant=CMM2026.0471———————
EChO-AgentSize=–2026.067163.6453.481.2981.8277.0674.3975
AudioJigsaw2026.0470.7———————
Qwen3-Omni ThinkingSize=30B2026.0669.967.35177.654.678.973.270.8
Qwen3-Omni Flash InstructSize=-2026.0669.866.758.374.281.878.464.670.8
OmniJigsawvariant=SMS2026.0469.5———————
SAR-LMSize=Agent2026.0669.352.756.381————
Step-Audio-R1Size=33B, Model Category=Open-source models2026.0669.15———————
OmniJigsawvariant=JMI2026.0469.1———————
Qwen-3-Omni-thinkingSize=30B2026.066966.6749.5179.9363.6474.3171.9562.5
AudioToolAgentSize=Agent2026.0668.861.851.977.672.776.67270.8
Step-Audio-R1.1Size=33B, Model Category=Open-source models2026.0668.75———————
Qwen-3-Omni-instructSize=30B2026.0668.758.7952.9178.2372.7377.0670.7370.83
Qwen3-Omni-30B2026.0468.5———————
Gemini 2.5 FlashSize=-2026.0668.46053.477.263.676.269.575
Qwen3-Omni Flash ThinkingSize=-2026.0668.165.551.975.563.676.265.970.8
MOSS-Audio-8B-ThinkingSize=8B, Model Category=Open-source models2026.0666.53———————
Qwen3-OmniReasoning=true, Model Size=30B, Zero-shot=true2026.0466.4———————
Qwen3-Omni-30B-A3B-InstructSize=30B, Model Category=Open-source models2026.0666.4———————
AudioToolAgent-OpenSource=Open Source2025.1065.961.2157.2869.0554.5569.2778.0566.67
Gemini 2.0 FlashSource=Closed Source2025.1065.661.2150.9772.1181.8272.4865.8570.83
Gemini 2.0 FlashZero-shot=true2026.0665.661.25172.181.872.565.9—
Gemini 2.0 FlashSize=–2026.0665.6—50.9772.1181.8272.4865.8570.83
MOSS-Audio-8B-InstructSize=8B, Model Category=Open-source models2026.0664.42———————
MOSS-Audio-4B-ThinkingSize=4B, Model Category=Open-source models2026.0663.91———————
GPT-4o AudioSource=Closed Source2025.1063.553.9450.9770.4163.6472.4862.275
GPT-4o AudioSize=-2026.0663.553.95170.463.672.562.275
Omni-R1Source=Closed Source2025.1063.467.351.564.345.570.264.670.8
MOSS-Audio-4B-InstructSize=4B, Model Category=Open-source models2026.0662.54———————
HumanOmniV22026.0461.8———————
MiMo-Audio-7BSize=7B, Model Category=Open-source models2026.0661.7———————
AudioToolAgent-OpenSize=Agent2026.0661.759.445.667.354.670.659.870.8
Kimi-AudioSize=7B, Model Category=Open-source models2026.0660.82———————
AF3 (think)Reasoning=true, Model Size=Teacher, Zero-shot=true2026.0460.1———————
GPT4o-AudioSize=–, Model Category=Closed-source models2026.0659.78———————
Omni-R12026.0459.7———————
AVRTReasoning=true, Model Size=7B, Zero-shot=true2026.0459.1———————
AudioGenie-ReasonerSize=Agent2026.0658.949.743.369.245.564.565.359.1
Audio Flamingo 3Size=8.3B, Setting=Base (reported)2026.0658.5———————
Audio Flamingo 3Size=7B, Model Category=Open-source models2026.0657.96———————
AVRTReasoning=true, Model Size=3B, Zero-shot=true2026.0457.3———————
Qwen2.5-OmniSource=Open Source2025.1056.758.7940.7859.8654.5561.9367.0758.33
Qwen2.5-OmniSize=7B, Model Category=Open-source models2026.0656.7———————
Qwen2.5-OmniModel Size=7B, Zero-shot=true2026.0656.758.840.859.954.661.967.1—
Qwen-2.5-OmniSize=7B2026.0656.758.7940.7859.8654.5561.9367.0758.33
Qwen2.5 OmniReasoning=false, Model Size=7B, Zero-shot=true2026.0456.5———————
SpectCount (Audio Flamingo 3)Size=8.3B, Trained Parameters=26.2M, Setting=SpectCount2026.0656.3———————
MiniCPM-o-4.5Size=9B, Model Category=Open-source models2026.0655.75———————
Qwen-2.5-OmniSize=3B2026.0653.853.9446.1253.7436.3660.0957.3258.33
Qwen2.5 OmniReasoning=false, Model Size=3B, Zero-shot=true2026.0453.7———————
Audio Flamingo 3Size=8.3B, Setting=Base (reproduced)2026.0652.9———————
v-SALMONN-o1Reasoning=true, Model Size=7B, Zero-shot=true2026.0451.2———————
Audio Flamingo NextSize=8B, Model Category=Open-source models2026.0651.01———————
GPT-4o mini AudioZero-shot=true2026.0650.638.835.958.845.560.157.3—
R1-AQASize=7B2026.0647.655.837.4499.1505050
Cap+DeepSeek-V3Model Size=671B, Zero-shot=true2026.0647.642.440.856.118.25045.1—
TinyGiantALMModel Size=1.5B, Zero-shot=true2026.0646.447.337.946.945.549.158.5—
SpectCount (Qwen2-Audio-Instruct)Size=8.4B, Trained Parameters=25.9M, Setting=SpectCount2026.0645.7———————
Baichuan-Omni-1.5Model Size=11B, Zero-shot=true2026.0640.741.23340.536.448.639—
Baichuan-Omni-1.5Size=11B2026.0640.741.2133.0140.4836.3648.6239.0241.67
Qwen2-Audio-InstructSize=8.4B, Setting=Base (reproduced)2026.0640.1———————
Audio ReasonerSource=Open Source2025.1036.843.6433.532.9945.4542.6631.7125
Audio-ReasonerSize=7B2026.0636.843.633.53345.542.731.725
Audio-ReasonerModel Size=8.4B, Zero-shot=true2026.0636.843.633.53345.542.731.7—
SALMONNSize=13B2026.0633.230.331.134.79.134.935.441.7
SALMONNModel Size=13B, Zero-shot=true2026.0633.230.331.134.79.134.935.4—
Audio-CoTSize=7B2026.0631.335.825.2349.130.730.537.5
Audio-CoTModel Size=8.4B, Zero-shot=true2026.0631.335.825.2349.130.730.5—
Qwen2-Audio-InstructSize=8.4B, Setting=Base (reported)2026.0630———————
Qwen2-AudioModel Size=8.4B, Zero-shot=true2026.063033.324.332.39.131.230.5—
Random GuessSize=-2026.0629.3229.3925.8831.482529.331.128.13
OpenOmniSize=8B2026.062720.6122.3335.3718.1827.0623.1725
GAMAModel Size=7B, Zero-shot=true2026.0626.529.124.327.927.324.828.1—
AnyGPT-chatSize=8B2026.0623.724.2419.4222.1127.2727.5226.8329.17
Flamingo 2Model Size=3B, Zero-shot=true2026.0621.924.917.520.818.226.623.2—
LTU-ASModel Size=7B, Zero-shot=true2026.06192014.119.19.120.628.1—