livekit / text-classification updated 5 months ago

turn-detector

Name: turn-detector
Author: livekit

An open-weights language model for contextually-aware end-of-utterance (EOU) detection in voice AI applications. The model predicts whether a user has finished speaking based on the semantic content of their transcribed speech, providing a critical complement to voice activity detection (VAD) systems.

Params

130 M

Context

8,192

Downloads 30d

690 K

Likes

133

Licence: other Not gated SAFETENSORS 14 languages View on Hugging Face ↗

Download history

daily snapshots · 10 days

692 K675 K

Jul 28Jul 31Aug 3Aug 6

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	f32	0.5 GB	1.1 GB	✅ Runs comfortably

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Specifications

Architecture: LlamaForCausalLM
Parameters: 130 M
Tensor type: F32
Context length: 8,192
Vocabulary: 49,154
Layers / heads: 30 / 9
Licence: other
First seen on the Hub: 2024-12-07
Base model: Qwen2.5-0.5B-Instruct
Training datasets: undisclosed
Added to our catalog: 2026-07-28

Family

Base model and the most-downloaded derivatives in the catalog.

Base Qwen2.5-0.5B-Instruct Qwen 6.6 M

Compare with

bge-reranker-v2-m3 19.2 M finbert 5.6 M bge-reranker-base 4.6 M Prompt-Guard-86M 4.5 M distilbert-base-uncased-finetuned-sst-2-english 3.7 M twitter-roberta-base-sentiment-latest 2.6 M

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB