Audio Reasoning on MMAR (test)
74.4Average ScoreGemini 2.5 Pro
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 2.5 Pro2026.04 | 74.4 | 67.3 | 56.8 | 82 | 100 | 84.9 | 80.5 | 66.7 | — | — | |
| Audio-DeepThinkerModel Category=Open-source Models Finetuned with RL2026.04 | 74 | 66.06 | 64.08 | 80.27 | 81.82 | 80.73 | 73.17 | 75 | — | — | |
| Audio-CogitoSize=30B-A3B2026.04 | 71.7 | 66.67 | 53.4 | 79.25 | 90.91 | 79.9 | 76.83 | 79.17 | 62.22 | 87 | |
| Qwen3-Omni-30B-A3B-InstructModel Category=Open-source Models2026.04 | 70.1 | 62.42 | 57.28 | 77.21 | 81.82 | 79.82 | 64.63 | 70.83 | — | — | |
| Gemini 2.5 Flash2026.04 | 68.4 | 60 | 53.4 | 77.2 | 63.6 | 76.2 | 69.5 | 75 | — | — | |
| Qwen3-Omni-ThinkingSize=30B-A3B2026.04 | 68 | 64.24 | 50 | 79.25 | 54.55 | 72.48 | 69.51 | 70.83 | 57.97 | 85 | |
| Qwen3-Omni-30B-A3B-ThinkingModel Category=Open-source Models2026.04 | 67.8 | 61.21 | 47.09 | 77.89 | 72.73 | 77.98 | 68.29 | 70.83 | — | — | |
| AudioMCQModel Category=Open-source Models Finetuned with RL2026.04 | 67.1 | — | — | — | — | — | — | — | — | — | |
| Gemini 2.5 FlashModel Category=Closed-source Models2026.04 | 67 | 58.18 | 47.09 | 76.53 | 63.64 | 77.52 | 76.83 | 54.17 | — | — | |
| Qwen3-Omni-InstructSize=30B-A3B2026.04 | 66.9 | 59.39 | 54.37 | 72.45 | 63.67 | 77.52 | 65.85 | 66.67 | — | — | |
| Gemini 2.0 Flash2026.04 | 65.6 | 61.2 | 51 | 72.1 | 81.8 | 72.5 | 65.9 | 70.8 | — | — | |
| Gemini 2.0 FlashModel Category=Closed-source Models2026.04 | 65.6 | 61.21 | 50.97 | 72.11 | 81.82 | 72.48 | 65.85 | 70.83 | — | — | |
| Audio-ThinkerModel Category=Open-source Models Finetuned with RL2026.04 | 65.3 | 68.48 | 53.88 | 64.29 | 72.73 | 71.56 | 73.17 | 66.67 | — | — | |
| AudioRouterBackbone=Qwen2.5-Omni2026.02 | 64.4 | 61.4 | 66.1 | 77.2 | 57.2 | 65.1 | 68.3 | 55.5 | — | — | |
| GPT-4o Audio2026.04 | 63.5 | 53.9 | 51 | 70.4 | 63.6 | 72.5 | 62.2 | 75 | — | — | |
| GPT-4o AudioModel Category=Closed-source Models2026.04 | 63.5 | 53.94 | 50.97 | 70.41 | 63.64 | 72.48 | 62.2 | 75 | — | — | |
| Omni-R1Size=8.4B2026.04 | 63.4 | 67.3 | 51.5 | 64.3 | 45.5 | 70.2 | 64.6 | 70.8 | — | — | |
| Omni-CLSTBackbone=Qwen2.5-Omni2026.02 | 63 | 58.2 | 64.7 | 76.5 | 54.9 | 63.2 | 67.4 | 56.1 | — | — | |
| AuTAgentBackbone=GPT-4o Audio2026.02 | 63 | 51.52 | 45.15 | 73.81 | — | — | — | — | — | — | |
| CESARModel Category=Open-source Models Finetuned with RL2026.04 | 62.7 | 66.06 | 55.83 | 62.24 | 63.64 | 67.43 | 60.98 | 66.67 | — | — | |
| Omni-R1Model Category=Open-source Models Finetuned with RL2026.04 | 61.2 | 63.6 | 51.5 | 62.2 | 72.7 | 65.1 | 62.2 | 70.8 | — | — | |
| Ke-Omni-RModel Category=Open-source Models Finetuned with RL2026.04 | 60.9 | 63.64 | 47.09 | 62.93 | 63.64 | 68.35 | 67.07 | 45.84 | — | — | |
| Tools w/ Desc.Backbone=GPT-4o Audio2026.02 | 60 | 50.91 | 43.2 | 72.79 | — | — | — | — | — | — | |
| Step-Audio-R1Size=33B2026.04 | 58.6 | 32.52 | 32.52 | 68.71 | 45.22 | 72.02 | 62.2 | 72.02 | 46.55 | 79 | |
| Audio Flamingo 3Model Category=Open-source Models2026.04 | 58.5 | — | — | — | — | — | — | — | — | — | |
| Kimi-AudioModel Category=Open-source Models2026.04 | 57.6 | — | — | — | — | — | — | — | — | — | |
| Audio Flamingo 3Size=7B2026.04 | 57.4 | 56.97 | 45.15 | 59.52 | 45.45 | 67.89 | 59.76 | 41.67 | 45.2 | 79 | |
| Qwen2.5-OmniSize=10.7B2026.04 | 56.7 | 58.8 | 40.8 | 59.9 | 54.5 | 61.9 | 67.1 | 58.3 | — | — | |
| Qwen-2.5-OmniModel Category=Open-source Models2026.04 | 56.7 | 58.79 | 40.78 | 59.86 | 54.55 | 61.93 | 67.07 | 58.33 | — | — | |
| Covo-AudioModel Category=Open-source Models2026.04 | 55.3 | 57.58 | 41.26 | 62.5 | 36.36 | 59.17 | 58.54 | 62.5 | — | — | |
| Vanilla (No-Tool)Backbone=GPT-4o Audio2026.02 | 55 | 43.64 | 35.44 | 67.01 | — | — | — | — | — | — | |
| Random SelectionBackbone=GPT-4o Audio2026.02 | 52.2 | 42.42 | 30.58 | 61.22 | — | — | — | — | — | — | |
| Qwen2.5-OmniBackbone=Qwen2.5-Omni2026.02 | 52.1 | 45.6 | 55.2 | 68.4 | 42.1 | 51.5 | 55.8 | 46.1 | — | — | |
| AudioRouterBackbone=Qwen2-Audio2026.02 | 51.8 | 46.5 | 54.6 | 67.1 | 44.1 | 50.9 | 54.3 | 45.4 | — | — | |
| R1-AQABackbone=Qwen2-Audio2026.02 | 51.3 | 46.2 | 54.1 | 66.8 | 43.5 | 50.2 | 53.7 | 44.6 | — | — | |
| GPT-4o mini Audio2026.04 | 50.6 | 38.8 | 35.9 | 58.8 | 45.5 | 60.1 | 57.3 | 60 | — | — | |
| GPT-4o mini AudioModel Category=Closed-source Models2026.04 | 50.6 | 38.79 | 35.92 | 58.84 | 45.45 | 60.09 | 57.32 | 50 | — | — | |
| All-Tools (Concat.)Backbone=GPT-4o Audio2026.02 | 49.7 | 38.79 | 29.61 | 64.29 | — | — | — | — | — | — | |
| AuTAgentBackbone=Qwen2-Audio-7B-Instruct2026.02 | 41.3 | 48.48 | 31.55 | 45.92 | — | — | — | — | — | — | |
| Audio-ReasonerSize=8.4B2026.04 | 40.5 | 42.42 | 32.52 | 42.52 | 45.13 | 48.62 | 30.49 | 29.17 | 28.4 | 68 | |
| Random SelectionBackbone=Qwen2-Audio-7B-Instruct2026.02 | 38.6 | 47.27 | 30.1 | 40.82 | — | — | — | — | — | — | |
| Tools w/ Desc.Backbone=Qwen2-Audio-7B-Instruct2026.02 | 37.5 | 40 | 27.67 | 40 | — | — | — | — | — | — | |
| Audio-ReasonerModel Category=Open-source Models Finetuned with RL2026.04 | 36.8 | 43.64 | 33.5 | 32.99 | 45.45 | 42.66 | 31.71 | 25 | — | — | |
| All-Tools (Concat.)Backbone=Qwen2-Audio-7B-Instruct2026.02 | 35.5 | 40 | 25.24 | 37.07 | — | — | — | — | — | — | |
| Vanilla (No-Tool)Backbone=Qwen2-Audio-7B-Instruct2026.02 | 35.1 | 37.58 | 22.33 | 38.78 | — | — | — | — | — | — | |
| SALMONNSize=7B2026.04 | 32.8 | 30.9 | 29.6 | 34.4 | 9.1 | 37.6 | 28.1 | 37.5 | — | — | |
| Audio-CoTSize=8.4B2026.04 | 31.3 | 35.8 | 25.2 | 34 | 9.1 | 30.7 | 30.5 | 37.5 | 19.85 | 62 | |
| Qwen2-AudioBackbone=Qwen2-Audio2026.02 | 30.4 | 28.5 | 30.2 | 35.4 | 26.8 | 31 | 32.5 | 28.4 | — | — | |
| Qwen2-AudioSize=8.4B2026.04 | 30.4 | 33.9 | 23.3 | 33 | 9.1 | 33 | 26.8 | 33.3 | — | — | |
| Qwen2-Audio-InstructSize=8.4B2026.04 | 30 | 33.3 | 24.3 | 32.3 | 9.1 | 31.2 | 30.5 | 25 | — | — | |
| MellowSize=167M2026.04 | 30 | 33.3 | 26.7 | 24.8 | 18.2 | 37.2 | 32.9 | 29.2 | 18.5 | 61 | |
| Qwen2-Audio-InstructModel Category=Open-source Models2026.04 | 30 | 33.33 | 24.27 | 32.31 | 9.09 | 31.19 | 30.49 | 25 | — | — | |
| Random Guess2026.04 | 29.32 | 29.39 | 25.88 | 31.48 | 25 | 29.3 | 31.1 | 28.13 | — | — | |
| Random Guess2026.02 | 29.3 | 29.4 | 25.9 | 31.5 | 25 | 29.3 | 31.1 | 28.1 | — | — | |
| Audio FlamingoSize=2.2B2026.04 | 26.6 | 32.7 | 21.8 | 24.8 | 18.2 | 30.3 | 24.4 | 25 | — | — | |
| Audio Flamingo 2Size=3B2026.04 | 21.9 | 24.9 | 17.5 | 20.8 | 18.2 | 26.6 | 23.2 | 8.3 | — | — | |
| Audio Flamingo 2Model Category=Open-source Models2026.04 | 21.9 | 24.85 | 17.48 | 20.75 | 18.18 | 26.61 | 23.17 | 8.33 | — | — |