NetEase Youdao's Open-Source Speech Models Top Hugging Face Trending

Around September 23, NetEase Youdao open-sourced two speech models, both topping Hugging Face trending in their categories: the streaming ASR model Confucius4-R2T2 and the Chinese-English simultaneous interpretation model Confucius4-T3PO, with weights available on Hugging Face. Multiple posters note the two can be combined into a fully open-source, locally deployable real-time speech translation stack.

Confirmed

• R2T2: 2B parameters, fine-tuned from Qwen3-ASR-1.7B; true streaming ASR with configurable decode chunks of 80ms–2s, 160ms latency, near-offline accuracy; append-only output (text is never revised once emitted), suited to live captions and voice agents; released under a custom NetEase license.
• T3PO: 14B-parameter Chinese-English simultaneous interpretation model that makes its own decisions such as when to segment; Apache 2.0 licensed.
• Both weights are on Hugging Face; combined they form a fully open-source, locally deployable interpretation stack.

Why it matters

• Real-time speech translation has largely relied on closed services; R2T2 + T3PO offer an end-to-end open alternative directly useful for captioning, interpretation, and voice-agent developers.
• R2T2's append-only design eliminates the "text jitter" common in streaming transcription — a key production-oriented tradeoff.

2026-09-23 ~ 2026-09-23 · 5 related posts

• NetEase Youdao open-sources 2B streaming ASR and 14B Chinese-English simultaneous translation models — AdinaYakup · 2026-09-23
• [source] NetEase Youdao's open-source ASR and translation models top Hugging Face trending — thetripathi58 · 2026-09-23
• NetEase Youdao open-sources Confucius4-R2T2 and T3PO, a full real-time voice AI stack — thetripathi58 · 2026-09-23

2 near-duplicate retellings: AdinaYakup · thetripathi58