Reasoning on BIG Bench Audio Text Modality
98.5AccuracyGemini-2.5-Flash
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-2.5-FlashModel Series=Gemini-2.5-Flash2026.03 | 98.5 | |
| Qwen3-A3B-InstructModel Series=Qwen3-A3B-Instruct, Modality=Base, Text-modality2026.03 | 97.72 | |
| X-OPDModel Series=Omni-A3B-Instruct, Training Paradigm=X-OPD2026.03 | 96.85 | |
| Gemini-3-Flash-PreviewModel Series=Gemini-3-Flash-Preview2026.03 | 95.28 | |
| Omni-A3B-InstructModel Series=Omni-A3B-Instruct, Training Paradigm=Base2026.03 | 92.1 | |
| GPT-4oModel Series=GPT-4o, Modality=Base, Text-modality2026.03 | 90.27 | |
| Omni-A3B-Instruct + Offline KDModel Series=Omni-A3B-Instruct, Training Paradigm=Offline KD2026.03 | 86.18 | |
| Omni-A3B-Instruct + GKD (Forward KL)Model Series=Omni-A3B-Instruct, Training Paradigm=GKD (Forward KL)2026.03 | 85.43 | |
| Omni-A3B-Instruct + SFTModel Series=Omni-A3B-Instruct, Training Paradigm=SFT2026.03 | 85.3 | |
| Qwen2.5-7B-InstructModel Series=Qwen2.5-7B-Instruct, Modality=Base, Text-modality2026.03 | 74.07 | |
| Omni-7B-InstructModel Series=Omni-7B-Instruct2026.03 | 66.49 | |
| Voxtral-Mini-3BModel Series=Voxtral-Mini-3B2026.03 | 59.8 | |
| Ministral-3-3BModel Series=Ministral-3-3B, Modality=Base, Text-modality2026.03 | 59.7 |