Speechocean762 (Realtime Word) - Word Total: leaderboard

Metric: Pearson correlation (-1 to 1) with expert word total ratings on valid outputs from the 2,500-utterance Speechocean762 test split; GPT-4o Realtime zero-shot, temperature 0.6, word prompt/run; 47.99% of the corresponding utterance/word/phoneme items unscored and excluded from correlation; higher is better. Source: arxiv.org. 1 models tracked.

Top models

#ModelScoreOverall rank
1GPT-4o Realtime (Speechocean762 snapshot unspecified)0.28

Interactive version: theaggregate.ai/benchmark?slug=speechocean762-realtime-word-word-total · How It Works · Data refreshed daily, snapshot 2026-10-12.