SEA-SpeechBench - Speaker Recognition (English Prompt): leaderboard
Metric: Macro-F1 (%; whether two clips come from the same speaker, on self-constructed same and different speaker pairs, refusals count as errors; 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: Around December 2026. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Flash | 58.01 |
| 2 | Qwen3 Omni 30B A3B Instruct | 56.15 |
| 3 | SeaLLMs-Audio-7B | 43.5 |
| 4 | Voxtral-Mini-3B-2507 | 42.84 |
| 5 | Qwen2-Audio-7B-Instruct | 41.93 |
| 6 | gemma-3n-E2B-it | 37.16 |
| 7 | Phi-4 Multimodal Instruct | 36.43 |
| 8 | Gemma 3n E4B (IT) | 33.01 |
Interactive version: theaggregate.ai/benchmark?slug=sea-speechbench-speaker-recognition-english-prompt · How It Works · Data refreshed daily, snapshot 2026-09-26.