Omni-DeepSearch: leaderboard
Current omni-modal benchmarks mainly evaluate models under settings where multiple modalities are provided simultaneously, while the ability to start from audio alone and actively search for cross-modal evidence remains underexplored. In this paper, we introduce Omni-DeepSearch, a benchmark for audio-driven omni-modal deep search. Given one or more audio clips and a related question, mod
Metric: Avg (self-reported). Source: benchmarklist.com. Saturation forecast: Around June 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 43.44 | #80 |
| 2 | Gemini 3 Flash | 20.47 | #99 |
| 3 | Gemini 2.5 Pro | 17.5 | #152 |
| 4 | Qwen 3.5 Omni Plus | 13.91 | #84 |
| 5 | MiMo-V2.5 | 11.72 | #196 |
| 6 | MiMo-V2-Omni | 9.69 | #253 |
| 7 | Qwen3.5 Omni Flash | 4.69 | #177 |
| 8 | Qwen3 Omni 30B A3B Instruct | 3.44 | #383 |
| 9 | Gemini 2.5 Flash Lite | 2.19 | #422 |
| 10 | Qwen2.5-Omni-7B | 1.09 | #624 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=omni-deepsearch · How It Works · Data refreshed daily, snapshot 2026-10-20.