Codec throughput, not model intelligence, may be the real constraint for on-device voice agents, and LiquidCodec targets exactly that layer with open weights and CPU-only speed. Teams building low-latency voice products on edge hardware — earbuds, glasses, vehicles — can benchmark it against Mimi or X-Codec 2 and deploy locally at Q4_0 precision without GPU dependency. Its open availability also makes it a practical drop-in for local speech pipelines.
#LiquidAI
6 stories · 1 in 7 days · 2 in 30 days
Stories, products and papers tagged "LiquidAI", newest first.
Nov 14
LiquidAI releases LiquidCodec, a 74M speech codec running fast on CPUs
Oct 28
Chinese contentLFM3-1.7B 量化版上线:8GB Mac 可本地运行智能体模型
LiquidAI 新出的 1.7B 智能体模型被量化到 1.1-1.8GB,8GB Mac 就能跑,想本地玩智能体的可以直接下 Q8_0 版本。
Oct 14
Liquid AI open-sources LFM2.5-24B-A2B with 2B active parameters
The 2B active-parameter design targets efficiency over raw capability, so the interesting question is how it trades benchmark rank for cheaper inference against larger dense models. Developers running edge or high-throughput deployments can evaluate the weights directly from the release link. The comparison with Qwen gives a concrete starting point for choosing between the two families.
Chinese contentLiquidAI LFM2.5-24B-A2B 本地对决 Gemma 4 26B 全胜
LiquidAI 的 24B 模型在 RTX 5090 上跑终端任务,8 场全赢 Gemma 4 26B,还快 3.1 倍,本地部署党可以看看。
Chinese contentLiquidAI 发布 LFM2.5-24B-A2B:可在本地设备运行私有智能体
LiquidAI 出了个能在自己设备上跑的 LFM2.5-24B-A2B,想做本地私有智能体的可以看看这篇介绍。
Oct 2