The Agent RaceJune 24, 2026via MarkTechPost
Gradium Launches stt-translate and s2s-translate, Real-Time Speech Translation Models Beating gpt-realtime-translate on Accuracy and Latency
Why it matters
A lesser-known player is challenging OpenAI's GPT-4 Realtime and Google's Gemini 3.5 Live on a critical multimodal capability—real-time speech-to-speech translation—with claims of superior accuracy-latency tradeoff and architectural efficiency (2-model cascade vs. 3-model pipeline).
Key signals
- Gradium released stt-translate and s2s-translate models
- Covers 5 languages (English, French, German, Spanish, Portuguese) across 20 language pairs
- Two-model cascade architecture vs. standard three-model pipeline
- Claims better accuracy-latency tradeoff than gpt-realtime-translate and gemini-3.5-live-translate
- Includes voice selection and voice cloning features
- Deployed via single duplex WebSocket connection
The hook
Gradium just beat OpenAI and Google on real-time speech translation. Here's the latency gap that matters.
Gradium released two real-time speech translation models, stt-translate and s2s-translate, covering English, French, German, Spanish, and Portuguese across 20 language pairs. The models collapse the standard three-model cascade into two, pairing single-pass transcription-and-translation with a Gradium TTS stage over one duplex WebSocket. Gradium reports a better accuracy-latency tradeoff than gpt-realtime-translate and gemini-3.5-live-translate, plus output voice selection and cloning.
The post Gradium Launches stt-translate and s2s-translate, Real-Time Speech Translation Models Beating gpt-realtime-translate on Accuracy and Latency appeared first on MarkTechPost.