SEA-SpeechBench - Temporal Localization (30-60 s): leaderboard
Metric: Span-overlap F1 (%; predict the start and end time at which a queried phrase is spoken in a recording of 30-60 seconds; English instruction prompt; blank where the model cannot take audio of that length). Source: arxiv.org. Saturation forecast: Around 2034. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen2.5-Omni-7B | 19.98 |
| 2 | GPT-4o Audio | 17.83 |
| 3 | Voxtral-Mini-3B-2507 | 9.87 |
| 4 | Gemma 3n E4B (IT) | 8.72 |
| 5 | gemma-3n-E2B-it | 8.27 |
| 6 | Gemini 2.5 Flash | 7.7 |
| 7 | Phi-4 Multimodal Instruct | 6.25 |
| 8 | Qwen3 Omni 30B A3B Instruct | 5.9 |
Interactive version: theaggregate.ai/benchmark?slug=sea-speechbench-temporal-localization-30-60-s · How It Works · Data refreshed daily, snapshot 2026-09-26.