Gradium Tops Voice Arena TTS Latency Benchmarks With 231ms Time-to-First-Audio
rohanpaul_ai · x · 2026-09-09
Voice Arena has published measured TTS latency results across 10 models — notable because the benchmark independently measures every model and publishes raw numbers, unlike self-reported scores.
Key results:
- Gradium ranked #1 with a p50 time-to-first-audio of just 231ms, the fastest of all tested models
- It also showed the tightest spread across p25/p50/p75, indicating the most consistent latency
The Gradium team notes TTFA is typically the first metric developers measure when evaluating voice models, making this result practically relevant for model selection.
Related event: Gradium tops Voice Arena TTS latency benchmark at 231ms(2 posts)→
More from Multimodal
- GPT Image 2.5 precision editing pulls off edits Image 2 couldn't handle — pbbakkum · 2026-09-09
- OpenAI Images 2.5 wows with a detailed 'history of painting 1825-1925' composition — DeryaTR_ · 2026-09-09
- Wan 2.2 I2V 14B custom LoRA Space trends on Hugging Face — Cockdaddyfuck · 2026-09-09
- OpenAI launches GPT-Image-2.5 Flare and Sunburst image models in the API — OpenAIDevs · 2026-09-09
- Suno partners with Believe and TuneCore to let AI-made songs reach streaming platforms — suno · 2026-09-09
- ChatGPT image gen loses to 9-month-old Nano Banana Pro on realism — flowersslop · 2026-09-09