List your company

Audio Reasoning on MMAR (test)

74.4Average ScoreGemini 2.5 Pro
19.833.97548.1562.325Feb 11, 2026Feb 22, 2026Mar 5, 2026Mar 17, 2026Mar 28, 2026Apr 8, 2026Apr 20, 2026

Evaluation results

MethodLinks
Gemini 2.5 Pro2026.0474.467.356.88210084.980.566.7——
Audio-DeepThinkerModel Category=Open-source Models Finetuned with RL2026.047466.0664.0880.2781.8280.7373.1775——
Audio-CogitoSize=30B-A3B2026.0471.766.6753.479.2590.9179.976.8379.1762.2287
Qwen3-Omni-30B-A3B-InstructModel Category=Open-source Models2026.0470.162.4257.2877.2181.8279.8264.6370.83——
Gemini 2.5 Flash2026.0468.46053.477.263.676.269.575——
Qwen3-Omni-ThinkingSize=30B-A3B2026.046864.245079.2554.5572.4869.5170.8357.9785
Qwen3-Omni-30B-A3B-ThinkingModel Category=Open-source Models2026.0467.861.2147.0977.8972.7377.9868.2970.83——
AudioMCQModel Category=Open-source Models Finetuned with RL2026.0467.1—————————
Gemini 2.5 FlashModel Category=Closed-source Models2026.046758.1847.0976.5363.6477.5276.8354.17——
Qwen3-Omni-InstructSize=30B-A3B2026.0466.959.3954.3772.4563.6777.5265.8566.67——
Gemini 2.0 Flash2026.0465.661.25172.181.872.565.970.8——
Gemini 2.0 FlashModel Category=Closed-source Models2026.0465.661.2150.9772.1181.8272.4865.8570.83——
Audio-ThinkerModel Category=Open-source Models Finetuned with RL2026.0465.368.4853.8864.2972.7371.5673.1766.67——
AudioRouterBackbone=Qwen2.5-Omni2026.0264.461.466.177.257.265.168.355.5——
GPT-4o Audio2026.0463.553.95170.463.672.562.275——
GPT-4o AudioModel Category=Closed-source Models2026.0463.553.9450.9770.4163.6472.4862.275——
Omni-R1Size=8.4B2026.0463.467.351.564.345.570.264.670.8——
Omni-CLSTBackbone=Qwen2.5-Omni2026.026358.264.776.554.963.267.456.1——
AuTAgentBackbone=GPT-4o Audio2026.026351.5245.1573.81——————
CESARModel Category=Open-source Models Finetuned with RL2026.0462.766.0655.8362.2463.6467.4360.9866.67——
Omni-R1Model Category=Open-source Models Finetuned with RL2026.0461.263.651.562.272.765.162.270.8——
Ke-Omni-RModel Category=Open-source Models Finetuned with RL2026.0460.963.6447.0962.9363.6468.3567.0745.84——
Tools w/ Desc.Backbone=GPT-4o Audio2026.026050.9143.272.79——————
Step-Audio-R1Size=33B2026.0458.632.5232.5268.7145.2272.0262.272.0246.5579
Audio Flamingo 3Model Category=Open-source Models2026.0458.5—————————
Kimi-AudioModel Category=Open-source Models2026.0457.6—————————
Audio Flamingo 3Size=7B2026.0457.456.9745.1559.5245.4567.8959.7641.6745.279
Qwen2.5-OmniSize=10.7B2026.0456.758.840.859.954.561.967.158.3——
Qwen-2.5-OmniModel Category=Open-source Models2026.0456.758.7940.7859.8654.5561.9367.0758.33——
Covo-AudioModel Category=Open-source Models2026.0455.357.5841.2662.536.3659.1758.5462.5——
Vanilla (No-Tool)Backbone=GPT-4o Audio2026.025543.6435.4467.01——————
Random SelectionBackbone=GPT-4o Audio2026.0252.242.4230.5861.22——————
Qwen2.5-OmniBackbone=Qwen2.5-Omni2026.0252.145.655.268.442.151.555.846.1——
AudioRouterBackbone=Qwen2-Audio2026.0251.846.554.667.144.150.954.345.4——
R1-AQABackbone=Qwen2-Audio2026.0251.346.254.166.843.550.253.744.6——
GPT-4o mini Audio2026.0450.638.835.958.845.560.157.360——
GPT-4o mini AudioModel Category=Closed-source Models2026.0450.638.7935.9258.8445.4560.0957.3250——
All-Tools (Concat.)Backbone=GPT-4o Audio2026.0249.738.7929.6164.29——————
AuTAgentBackbone=Qwen2-Audio-7B-Instruct2026.0241.348.4831.5545.92——————
Audio-ReasonerSize=8.4B2026.0440.542.4232.5242.5245.1348.6230.4929.1728.468
Random SelectionBackbone=Qwen2-Audio-7B-Instruct2026.0238.647.2730.140.82——————
Tools w/ Desc.Backbone=Qwen2-Audio-7B-Instruct2026.0237.54027.6740——————
Audio-ReasonerModel Category=Open-source Models Finetuned with RL2026.0436.843.6433.532.9945.4542.6631.7125——
All-Tools (Concat.)Backbone=Qwen2-Audio-7B-Instruct2026.0235.54025.2437.07——————
Vanilla (No-Tool)Backbone=Qwen2-Audio-7B-Instruct2026.0235.137.5822.3338.78——————
SALMONNSize=7B2026.0432.830.929.634.49.137.628.137.5——
Audio-CoTSize=8.4B2026.0431.335.825.2349.130.730.537.519.8562
Qwen2-AudioBackbone=Qwen2-Audio2026.0230.428.530.235.426.83132.528.4——
Qwen2-AudioSize=8.4B2026.0430.433.923.3339.13326.833.3——
Qwen2-Audio-InstructSize=8.4B2026.043033.324.332.39.131.230.525——
MellowSize=167M2026.043033.326.724.818.237.232.929.218.561
Qwen2-Audio-InstructModel Category=Open-source Models2026.043033.3324.2732.319.0931.1930.4925——
Random Guess2026.0429.3229.3925.8831.482529.331.128.13——
Random Guess2026.0229.329.425.931.52529.331.128.1——
Audio FlamingoSize=2.2B2026.0426.632.721.824.818.230.324.425——
Audio Flamingo 2Size=3B2026.0421.924.917.520.818.226.623.28.3——
Audio Flamingo 2Model Category=Open-source Models2026.0421.924.8517.4820.7518.1826.6123.178.33——