SEABED - Dialect Identification (Multiple-Choice): leaderboard
SEABED (Rajgarhia et al., 2026) is an audio-first reasoning benchmark on Southeast Asian speech built from openly available corpora. In this task the model hears four concatenated utterances in different Indonesian or Thai lects and answers questions about which lect was spoken where; multiple-choice items with ten options are scored by exact match, and the score is accuracy in percent, higher is better.
Metric: Accuracy (%; 504 dialect and language identification over four concatenated Indonesian or Thai lects items; multiple-choice items scored by exact match against the shuffled gold option; audio at 16 kHz). Source: arxiv.org. Saturation forecast: Around October 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.5 Flash | 41.5 | #98 |
| 2 | Gemini 3.6 Flash | 39.3 | #72 |
| 3 | Gemini 2.5 Pro | 37.5 | #262 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=seabed-dialect-identification-multiple-choice · How It Works · Data refreshed daily, snapshot 2026-11-12.