SEA-SpeechBench - Gender Recognition (English Prompt): leaderboard
Metric: Macro-F1 (%; speaker gender from speech, labels canonicalised by bilingual string matching, per-dataset scores averaged; short clips of at most 30 s from curated public Southeast Asian speech corpora, up to 1,000 sampled items per dataset; English instruction prompt). Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen3 Omni 30B A3B Instruct | 95.78 |
| 2 | Gemini 2.5 Flash | 92.54 |
| 3 | Qwen2-Audio-7B-Instruct | 91.89 |
| 4 | Qwen2.5-Omni-7B | 65.03 |
| 5 | GPT-4o Audio | 59.4 |
| 6 | Phi-4 Multimodal Instruct | 49.36 |
| 7 | Voxtral-Mini-3B-2507 | 39.78 |
| 8 | SeaLLMs-Audio-7B | 37.07 |
| 9 | Gemma 3n E4B (IT) | 35.41 |
Interactive version: theaggregate.ai/benchmark?slug=sea-speechbench-gender-recognition-english-prompt · How It Works · Data refreshed daily, snapshot 2026-09-26.