SpeechParaling-Bench (English) - Dynamic Variation: leaderboard

SpeechParaling-Bench (2026) tests paralinguistic-aware speech generation by audio-language models; a Gemini 3 Pro judge compares each spoken response with the response of a fixed baseline model, ties counting half, and the baseline itself scores its strength-weighted result against the others. This board covers the dynamic variation task (change pitch, speed or volume within one utterance) in the English subset, against the Gemini 2.5 Flash Audio baseline. Pairwise score in percent; higher is better.

Metric: Pairwise score (%) on the dynamic variation task (change a feature such as pitch or speed within one utterance), English subset: a Gemini 3 Pro judge compares each spoken response with the fixed baseline's (Gemini 2.5 Flash Audio; ties count half), and the baseline's own row is its strength-weighted average against the candidates (the paper's Eq. 3); higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 5 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash Audio61.08
2GPT-Audio (2025-08-28)52.92
3Doubao Realtime Voice Model22.5
4Qwen3-Omni-Flash-2025-12-017.92
5Qwen3 Omni Realtime5

Interactive version: theaggregate.ai/benchmark?slug=speechparaling-bench-english-dynamic-variation · How It Works · Data refreshed daily, snapshot 2026-11-21.