Audio-Visual Understanding on DailyOmni
82.71Average ScoreOmniAgent
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OmniAgentModality=A+V2025.12 | 82.71 | 81.05 | 80.67 | 83.36 | 86.86 | 80.83 | 83.21 | 80.37 | 85.45 | |
| D-ORCAModel Parameters=8B2026.02 | 78.5 | — | — | — | — | — | — | — | — | |
| OmniSIFTBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.02 | 73.2 | 66.7 | 70.2 | 83.1 | 78.9 | 69.9 | 79.4 | — | — | |
| Gemini 2.5-FlashModality=A+V, Chain-of-Thought (CoT)=true2025.12 | 72.7 | — | — | — | — | — | — | — | — | |
| OmniSIFTBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.02 | 72.5 | 66.7 | 68.9 | 82.5 | 77.7 | 71 | 76.3 | — | — | |
| Full TokensBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=1002026.02 | 72.2 | 66.7 | 70.6 | 79.2 | 76.6 | 69.9 | 77.1 | — | — | |
| Qwen3-Omni-30BModality=A+V2025.12 | 72.08 | 65.32 | 61.9 | 82.67 | 85.92 | 69.47 | 79.25 | 71.28 | 74.29 | |
| video-SALMONN 2+Model Parameters=7B2026.02 | 71.8 | — | — | — | — | — | — | — | — | |
| XGC-AVisModality=A+V2025.12 | 71.5 | 64.4 | 63.5 | 85.1 | 82.3 | 68.4 | 77.1 | 71.6 | 71.5 | |
| AVoCaDOModel Parameters=7B2026.02 | 69.8 | — | — | — | — | — | — | — | — | |
| Qwen3-Omni-InstructModel Parameters=30B-A3B2026.02 | 69.8 | — | — | — | — | — | — | — | — | |
| DyCokeBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.02 | 67.9 | 61.4 | 63.9 | 77.9 | 75.4 | 63.7 | 74.8 | — | — | |
| Gemini 2.0-FlashModality=A+V2025.12 | 67.84 | 63.72 | 62.18 | 76.62 | 75.43 | 63.73 | 73.28 | 67.23 | 68.55 | |
| OmniZipBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.02 | 67.7 | 63.7 | 63 | 77.3 | 76.6 | 59.1 | 74.8 | — | — | |
| Full TokensBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=1002026.02 | 67 | 60.1 | 62.2 | 78.6 | 74.9 | 62.2 | 74.8 | — | — | |
| OmniVinciModel Parameters=7B2026.02 | 66.5 | — | — | — | — | — | — | — | — | |
| RandomBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.02 | 66.3 | 58.5 | 61.8 | 77.9 | 73.7 | 63.2 | 74 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.02 | 66.2 | 61.8 | 59.7 | 75.3 | 75.4 | 60.6 | 74 | — | — | |
| OmniSIFTBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.02 | 65.3 | 57.8 | 58.8 | 77.3 | 73.7 | 64.8 | 69.5 | — | — | |
| RandomBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.02 | 65.2 | 61.1 | 56.7 | 78.6 | 71.4 | 60.1 | 73.3 | — | — | |
| Qwen3-VL-PlusModality=V2025.12 | 65 | 66.99 | 51.68 | 71.43 | 68.57 | 61.66 | 77.1 | 63.68 | 66.55 | |
| DyCokeBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.02 | 64.7 | 57.2 | 56.7 | 80 | 74.3 | 61.1 | 71 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.02 | 64.7 | 60.5 | 56.7 | 76.6 | 72 | 59.6 | 72.5 | — | — | |
| OmniSIFTBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.02 | 64.7 | 58.5 | 59.7 | 75.3 | 73.7 | 60.6 | 70.2 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.02 | 64.2 | 57.8 | 55 | 75.3 | 70.3 | 58.5 | 70 | — | — | |
| Full TokensBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=100%, Frame Budgets=1282026.05 | 64.16 | — | 50.84 | 78.57 | 78.86 | 56.99 | 71.76 | — | — | |
| DyCokeBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.02 | 63.2 | 53.9 | 52.5 | 79.2 | 76 | 60.1 | 72.5 | — | — | |
| ARC-Qwen-Video-NarratorModel Parameters=7B2026.02 | 63.2 | — | — | — | — | — | — | — | — | |
| RandomBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.02 | 62.9 | 55.9 | 54.2 | 76 | 74.3 | 60.1 | 68.7 | — | — | |
| Full TokensBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=100%, Frame Budgets=1282026.05 | 62.82 | — | 53.78 | 79.22 | 74.29 | 55.44 | 70.99 | — | — | |
| Qwen2.5-OmniModel Parameters=7B2026.02 | 62.7 | — | — | — | — | — | — | — | — | |
| Full TokensModel=Qwen2.5-Omni-7B, Retained Ratio (%)=1002026.07 | 61.9 | 57.52 | 48.32 | 75.32 | 73.14 | 59.59 | 69.47 | — | — | |
| Daily-OmniModality=A+V2025.12 | 61.82 | 53.92 | 51.68 | 78.57 | 71.43 | 60.1 | 68.7 | 63.99 | 59.27 | |
| DyCokeBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.02 | 61.7 | 52.6 | 54.6 | 74.7 | 74.3 | 58 | 69.5 | — | — | |
| RandomBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.02 | 61.2 | 53.3 | 54.2 | 76.6 | 72 | 55.4 | 67.9 | — | — | |
| OmniSelectBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=45%, Frame Budgets=1282026.05 | 60.65 | — | 46.64 | 76.62 | 76.57 | 53.37 | 72.52 | — | — | |
| OmniFocusModel=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.07 | 60.32 | 52.61 | 48.32 | 78.57 | 72.57 | 55.96 | 68.7 | — | — | |
| Full TokensModel=Qwen2.5-Omni-3B, Retained Ratio (%)=1002026.07 | 60.15 | 53.59 | 51.68 | 74.68 | 70.86 | 53.37 | 69.47 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=45%, Frame Budgets=1282026.05 | 59.98 | — | 46.22 | 77.27 | 76.57 | 53.89 | 70.23 | — | — | |
| DyCoke (V&A)Backbone=Qwen2.5-Omni-7B, Retained Ratio (%)=50%, Frame Budgets=1282026.05 | 59.48 | — | 44.96 | 77.92 | 76.57 | 50.26 | 70.23 | — | — | |
| OmniFocusModel=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.07 | 59.4 | 52.29 | 46.64 | 75.32 | 72.57 | 54.92 | 69.47 | — | — | |
| OmniZipModel=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.07 | 59.23 | 53.59 | 43.28 | 79.22 | 73.71 | 52.85 | 67.94 | — | — | |
| DVDModality=V2025.12 | 59.22 | 58.45 | 49.32 | 70.37 | 63.24 | 57.12 | 57.47 | 56.31 | 62.41 | |
| RandomModel=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.07 | 58.23 | 52.61 | 42.86 | 72.08 | 72.57 | 53.89 | 70.23 | — | — | |
| OmniSelectBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=45%, Frame Budgets=1282026.05 | 58.06 | — | 52.1 | 74.03 | 69.14 | 50.78 | 65.65 | — | — | |
| RandomModel=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.07 | 57.98 | 53.59 | 40.76 | 75.32 | 68.57 | 54.92 | 69.47 | — | — | |
| OmniZipModel=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.07 | 57.73 | 51.63 | 42.02 | 75.97 | 72.57 | 51.81 | 67.94 | — | — | |
| DyCokeModel=Qwen2.5-Omni-7B, Retained Ratio (%)=352026.07 | 57.57 | 56.64 | 43.02 | 68.48 | 71.84 | 57.25 | 62.77 | — | — | |
| OmniSelectBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=30%, Frame Budgets=1282026.05 | 57.39 | — | 46.22 | 74.68 | 73.14 | 48.19 | 67.18 | — | — | |
| OmniFocusModel=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.07 | 57.23 | 48.04 | 47.9 | 75.97 | 69.14 | 51.3 | 66.41 | — | — | |
| RandomBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=55%, Frame Budgets=1282026.05 | 56.89 | — | 41.48 | 72.73 | 71.43 | 52.33 | 70.99 | — | — | |
| GPT-4oModality=V2025.12 | 56.47 | 52.61 | 47.9 | 66.23 | 66.29 | 52.33 | 62.6 | 55.64 | 57.45 | |
| DyCokeModel=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.07 | 56.15 | 55.68 | 48.22 | 73.87 | 67.24 | 54.49 | 57.48 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=45%, Frame Budgets=1282026.05 | 56.14 | — | 46.64 | 76.62 | 69.71 | 47.15 | 64.12 | — | — | |
| OmniFocusModel=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.07 | 56.14 | 48.04 | 47.9 | 72.73 | 66.86 | 50.78 | 64.12 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-7B, Retained Ratio (%)=30%, Frame Budgets=1282026.05 | 55.97 | — | 43.7 | 76.62 | 73.71 | 47.15 | 63.36 | — | — | |
| DyCoke (V&A)Backbone=Qwen2.5-Omni-3B, Retained Ratio (%)=50%, Frame Budgets=1282026.05 | 55.89 | — | 47.48 | 72.73 | 65.14 | 48.19 | 65.65 | — | — | |
| OmniZipModel=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.07 | 55.85 | 47.71 | 44.54 | 75.32 | 66.67 | 52.33 | 63.36 | — | — | |
| Ola-7B + AVATARReinforcement Learning Strategy=AVATAR2025.08 | 55.7 | — | — | — | — | — | — | — | — | |
| OmniZipModel=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.07 | 55.52 | 48.37 | 46.22 | 74.68 | 66.09 | 49.22 | 61.83 | — | — | |
| RandomModel=Qwen2.5-Omni-3B, Retained Ratio (%)=352026.07 | 54.97 | 46.41 | 44.96 | 74.03 | 64 | 51.3 | 64.12 | — | — | |
| DyCokeModel=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.07 | 54.65 | 53.98 | 46.92 | 71.77 | 64.24 | 49.99 | 59.58 | — | — | |
| OmniSelectBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=30%, Frame Budgets=1282026.05 | 54.39 | — | 48.32 | 72.08 | 65.14 | 46.63 | 62.6 | — | — | |
| DyCokeModel=Qwen2.5-Omni-7B, Retained Ratio (%)=252026.07 | 54.37 | 55.54 | 38.82 | 65.88 | 68.04 | 59.35 | 55.57 | — | — | |
| Ola-7B + GRPOReinforcement Learning Strategy=GRPO2025.08 | 54.1 | — | — | — | — | — | — | — | — | |
| Qwen 2.5 Omni + MoD-DPO++base_model=Qwen 2.5 Omni, method=MoD-DPO++2026.03 | 53.82 | — | — | — | — | — | — | 53.45 | 54.18 | |
| AV-ReasonerReinforcement Learning Strategy=N/A2025.08 | 53.8 | — | — | — | — | — | — | — | — | |
| RandomModel=Qwen2.5-Omni-3B, Retained Ratio (%)=252026.07 | 53.72 | 45.1 | 43.7 | 71.43 | 64.57 | 48.7 | 64.12 | — | — | |
| RandomBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=55%, Frame Budgets=1282026.05 | 53.55 | — | 45.38 | 71.43 | 63.43 | 47.67 | 64.12 | — | — | |
| OmniZipBackbone=Qwen2.5-Omni-3B, Retained Ratio (%)=30%, Frame Budgets=1282026.05 | 53.05 | — | 42.86 | 68.18 | 68 | 46.63 | 59.54 | — | — | |
| Qwen 2.5 Omni + MoD-DPObase_model=Qwen 2.5 Omni, method=MoD-DPO2026.03 | 53 | — | — | — | — | — | — | 52.45 | 53.54 | |
| Ola-7BReinforcement Learning Strategy=Baseline2025.08 | 52.3 | — | — | — | — | — | — | — | — | |
| Qwen 2.5 Omni + DPObase_model=Qwen 2.5 Omni, method=DPO2026.03 | 51.44 | — | — | — | — | — | — | 50.95 | 51.93 | |
| Ola-7BModality=A+V2025.12 | 50.71 | 43.46 | 40.34 | 63.64 | 69.71 | 40.41 | 61.07 | 51.47 | 49.82 | |
| Qwen 2.5 Omni + OmniDPObase_model=Qwen 2.5 Omni, method=OmniDPO2026.03 | 50.07 | — | — | — | — | — | — | 49.24 | 50.89 | |
| Gemini 2.5-Flash + RAG + AgentModality=A+V, RAG=true, Agentic Reasoning=true2026.02 | 48.83 | 53.23 | 26.92 | 59.09 | 60 | 46.65 | 34.38 | — | — | |
| HumanOmniReinforcement Learning Strategy=N/A2025.08 | 47.6 | — | — | — | — | — | — | — | — | |
| Qwen2.5-Omni-7BModality=A+V2025.12 | 47.45 | 40.52 | 44.12 | 57.79 | 61.71 | 38.86 | 51.15 | 46.68 | 48.36 | |
| Qwen 2.5 Omnibase_model=Qwen 2.5 Omni2026.03 | 47.34 | — | — | — | — | — | — | 46.74 | 48.46 | |
| Qwen2.5-Omni + AVATARReinforcement Learning Strategy=AVATAR2025.08 | 47 | — | — | — | — | — | — | — | — | |
| Omni-R1Reinforcement Learning Strategy=N/A2025.08 | 46.8 | — | — | — | — | — | — | — | — | |
| Gemini 2.5-Flash + RAGModality=A+V, RAG=true2026.02 | 46.48 | 41.94 | 38.46 | 54.55 | 55.56 | 45.65 | 39.39 | — | — | |
| EcholnkReinforcement Learning Strategy=N/A2025.08 | 46.2 | — | — | — | — | — | — | — | — | |
| Qwen2.5-Omni-7B + RAG + Agent + RLModality=A+V, RAG=true, Agentic Reasoning=true, RL Optimization=true2026.02 | 44.92 | 59.09 | 27.45 | 45.45 | 45.95 | 51.22 | 32.14 | — | — | |
| Qwen2.5-Omni + GRPOReinforcement Learning Strategy=GRPO2025.08 | 44.8 | — | — | — | — | — | — | — | — | |
| Gemini 2.5-FlashModality=A+V2026.02 | 44.53 | 37.1 | 34.63 | 56.82 | 51.11 | 50 | 34.38 | — | — | |
| Qwen2.5-OmniReinforcement Learning Strategy=Baseline2025.08 | 44 | — | — | — | — | — | — | — | — | |
| Qwen2.5-Omni-7B + RAG + AgentModality=A+V, RAG=true, Agentic Reasoning=true2026.02 | 42.19 | 33.33 | 45.1 | 48.48 | 35.14 | 56.1 | 39.29 | — | — | |
| Qwen2.5-Omni-3B + RAG + Agent + RLModality=A+V, RAG=true, Agentic Reasoning=true, RL Optimization=true2026.02 | 40.09 | 40.54 | 39.22 | 36.36 | 51.35 | 41.46 | 28.57 | — | — | |
| Qwen2.5-Omni-7B + RAGModality=A+V, RAG=true2026.02 | 39.84 | 33.33 | 37.25 | 42.42 | 48.65 | 46.34 | 35.71 | — | — | |
| Qwen2.5-Omni-3B + RAG + AgentModality=A+V, RAG=true, Agentic Reasoning=true2026.02 | 37.11 | 25.76 | 37.25 | 42.42 | 51.35 | 41.46 | 32.14 | — | — | |
| MiniCPM-O 2.6 + MoD-DPO++base_model=MiniCPM-O 2.6, method=MoD-DPO++2026.03 | 36.6 | — | — | — | — | — | — | 43.33 | 29.87 | |
| Qwen2.5-Omni-7BModality=A+V2026.02 | 36.33 | 51.52 | 29.41 | 21.21 | 32.43 | 41.46 | 28.57 | — | — | |
| Qwen2.5-Omni-3B + RAGModality=A+V, RAG=true2026.02 | 35.94 | 22.73 | 27.45 | 48.48 | 37.84 | 60.98 | 28.57 | — | — | |
| MiniCPM-O 2.6 + MoD-DPObase_model=MiniCPM-O 2.6, method=MoD-DPO2026.03 | 35.71 | — | — | — | — | — | — | 42.86 | 28.56 | |
| VideoLLaMA2-7BModality=A+V2025.12 | 35.17 | 31.7 | 35.71 | 40.91 | 34.29 | 35.75 | 35.88 | 38.02 | 31.82 | |
| MiniCPM-O 2.6 + OmniDPObase_model=MiniCPM-O 2.6, method=OmniDPO2026.03 | 34.33 | — | — | — | — | — | — | 40.68 | 27.98 | |
| MiniCPM-O 2.6 + DPObase_model=MiniCPM-O 2.6, method=DPO2026.03 | 34.13 | — | — | — | — | — | — | 39.97 | 28.28 | |
| Qwen2.5-Omni-3BModality=A+V2026.02 | 32.03 | 25.76 | 39.22 | 33.33 | 29.73 | 41.46 | 21.43 | — | — | |
| MiniCPM-O 2.6base_model=MiniCPM-O 2.62026.03 | 30.55 | — | — | — | — | — | — | 38.02 | 23.09 |