AI BriefingTYGER Released: Open-Weights ASR with Full-Duplex Turn Control
·2026.10.15 18:04
Key point
The model achieves 250 ms median endpoint latency and integrates with Gemma 4 E4B and VoxCPM2 for voice agents.
Details
TYGER is an open-weights Automatic Speech Recognition (ASR) model that integrates full-duplex turn control, allowing it to detect when a user finishes speaking, interrupts, or provides backchannel cues like "mm-hm".
Performance and Architecture
- Latency: Achieves a 250 ms median endpoint latency with 80 ms frames and no lookahead.
- Voice Agent Stack: The full voice agent, tyger-s2s, combines TYGER with Gemma 4 E4B and VoxCPM2.
- Hardware Requirements: Runs on a single 32 GB GPU, supporting up to 4 concurrent conversations.
Availability
The model is currently available for English only. Code and weights are hosted on Hugging Face and GitHub under the nineninesix organization.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Wispr Unveils 'Canto', a Real-Time Transcription-Specialized Voice Model Achieving Lower WER Than Competitors via GRPO Training
Meta Releases Real-Time Multilingual ASR Model 'Muse Voice Transcribe'… Tops Benchmarks with 3.1% WER