Unverified50% confidenceFactExact time

Silero VAD等现代神经网络VAD模型参数量仅约1MB，在20ms帧级粒度上实现实时检测

1

Sources

50%

Confidence

Long-term

Relevance

7/22/2026

First Seen

Sources

Moonshine: A Low-Latency Speech Engine and Open-Source Powerhouse for Building Voice Agents

Related Claims

Unverified测试者用双4090（48GB显存）工作站运行20.7B稠密模型，每秒生成27.6个Token，仅比矮星略快，但参数量小10倍以上64% similar UnverifiedViT-L模型约3亿参数，checkpoint达1213MB，CPU上处理每张图像需4.5秒63% similar UnverifiedNeuralink的N1芯片理论上可同时记录1024个通道的神经信号，数据传输率可达约1Mbps59% similar UnverifiedTTM 模型参数量仅约 0.8M（百万级），文件大小约三兆字节，无需 GPU，在普通 CPU 上就能运行58% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/585917

API

curl https://kongchang.com/api/v1/knowledge/claims/585917

MCP

get_claim(id=585917)