Paralinguistic tasks on VoxParadox
30.85GT AccuracyAudio Flamingo 2 (AF2)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Audio Flamingo 2 (AF2)Model category=Open-Source Audio LLMs2026.05 | 30.85 | 29.8 | -1.05 | |
| Gemini 2.5 FlashModel category=Closed-Source Models2026.05 | 24.7 | 60.45 | 35.75 | |
| MiMo-AudioModel category=Open-Source Audio LLMs2026.05 | 19.6 | 68.95 | 49.35 | |
| Kimi-AudioModel category=Open-Source Audio LLMs2026.05 | 19 | 69.55 | 50.55 | |
| Step-Audio-R1Model category=Open-Source Audio LLMs2026.05 | 17.45 | 69.25 | 51.8 | |
| Audio Flamingo 3 (AF3)Model category=Open-Source Audio LLMs2026.05 | 17.4 | 68.5 | 51.1 | |
| Qwen2-AudioModel category=Open-Source Audio LLMs2026.05 | 14.85 | 70.25 | 55.4 | |
| Qwen3-OmniModel category=Open-Source Omni LLMs2026.05 | 10.6 | 80.65 | 70.05 | |
| GPT-4o AudioModel category=Closed-Source Models2026.05 | 8.6 | 81.55 | 72.95 | |
| Qwen2.5-OmniModel category=Open-Source Omni LLMs2026.05 | 7.95 | 75.2 | 67.25 | |
| VITA-AudioModel category=Open-Source Audio LLMs2026.05 | 6.85 | 75.45 | 68.6 | |
| SALMONNModel category=Open-Source Audio LLMs2026.05 | 6.1 | 22.45 | 16.35 |