SEA-SpeechBench - Emotion Recognition (English Prompt): leaderboard
Metric: Judge-based accuracy (%; closed nine-class emotion label from speech, free-form answers mapped to labels and judged by Gemma-3-27B-Instruct, 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: Around 2034. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen2-Audio-7B-Instruct | 24.47 |
| 2 | Phi-4 Multimodal Instruct | 20.87 |
| 3 | Gemini 2.5 Flash | 19.5 |
| 4 | GPT-4o Audio | 17 |
| 5 | Qwen2.5-Omni-7B | 16.33 |
| 6 | Qwen3 Omni 30B A3B Instruct | 15.94 |
| 7 | Gemma 3n E4B (IT) | 12.46 |
| 8 | SeaLLMs-Audio-7B | 12.34 |
| 9 | gemma-3n-E2B-it | 12.21 |
| 10 | Voxtral-Mini-3B-2507 | 10.62 |
Interactive version: theaggregate.ai/benchmark?slug=sea-speechbench-emotion-recognition-english-prompt · How It Works · Data refreshed daily, snapshot 2026-09-26.