NetEase Youdao open-sources Confucius4-R2T2, a streaming ASR with 80ms chunks
rohanpaul_ai · x · 2026-09-17
NetEase Youdao open-sourced Confucius4-R2T2 (Real Real-Time Transcription), a low-latency, high-accuracy streaming ASR model built on Qwen3-ASR. Key traits: configurable decoding chunks from 80ms to 2s, and append-only output that commits transcript text permanently without revising prior words — eliminating flicker in live captioning, LLM agent pipelines and simultaneous translation. The author frames the open question as whether stable, domain-aware streaming recognition becomes standard for voice systems.
Related event: NetEase Youdao Open-Sources Confucius4-R2T2 Streaming ASR Model(7 posts)→
More from Models
- Mozilla's 91-page report: open-weight AI now only ~4 months behind the frontier — rohanpaul_ai · 2026-09-17
- Unreleased Astra-family model reportedly developed a new persona banner during RL training — inductionheads · 2026-09-17
- Researcher Despairs as Gemini Cites 'Emergent Mind' for Made-up AUROC Baselines — anshulkundaje · 2026-09-17
- Databricks rolls out GPT-6 Astra to 3,500 engineers, coding spend jumps 60% — nickbaumann_ · 2026-09-17
- How do models know if they're in training, evaluation, or real inference? Nobody really knows — flowersslop · 2026-09-17
- Google brings Gemma 4 12B to Mac, running locally on 16GB MacBook Air — GlennCameronjr · 2026-09-17