TAU-HOME.COM stories tagged OnDeviceAI.
Liquid AI has launched LFM2.5-24B-A2B, an open-weight MoE model activating approximately 2B of its 24B total parameters per token to deliver fast local inferenc
Liquid AI has released LFM2.5-24B-A2B, an open-weight MoE reasoning model activating 2.45B of its 24.0B total parameters. Designed for on-device agents, it feat
OpenBMB has officially released MiniCPM-V 4.7, an open-source multimodal model series featuring Lookahead Expert Scheduling (LES). The 35B-A3B model delivers on
Tencent has unveiled Hy-MT2, an ultra-compact 440MB open-source multilingual AI translation model capable of running entirely offline on smartphones and edge de
The Google AI Edge team has released Google AI Edge Foresight for Mac, an experimental on-device meeting companion built with Flutter that pairs EmbeddingGemma
Microsoft and NVIDIA have unveiled the Surface Laptop Ultra with RTX Spark, offering up to 128GB of unified memory to run 120B AI models locally on top-tier mod
Ghost AI has unveiled Core, a personal AI computer running local models 24/7 on-device. Powered by NVIDIA RTX PRO 4000 Blackwell SFF, it hits 87 tok/s on Qwen-3
Google DeepMind has released EmbeddingGemma 2, an Apache 2.0 740M multimodal embedding model mapping text, audio, and vision into a unified 768-dim vector space
Cactus Compute has officially unveiled Whistle, a 16.9MB single-file speech-to-text (STT) model that runs entirely on CPUs. Operating 9x smaller and 6x faster t
The FlashML research team has unveiled FreeVideo, an open-source framework combining MiniMax H3 and Video DeltaNet to enable local video generation on consumer
Developer atemoya has released v2 of a Qwen3.5-0.8B-based Japanese SFT model on Hugging Face. The update focuses on more natural Japanese output rather than kno
FreeVideo enables on-device video and asset generation on consumer PCs. Paired with FreeToken, it automates the pipeline from storyline planning to video render
Google has updated the Antigravity SDK to support local offline execution via Google AI Edge's LiteRT engine, running Gemma 4 26B A4B with zero API costs.
An in-depth look at Edge0, an open-source on-device AI framework designed to run 35B parameter LLMs using just 1–2.5GB peak memory on Apple Silicon without clou
OpenBMB, Tsinghua University, and ModelBest have released MiniCPM5-2B, an open-source 2.52B on-device model running offline agents and 128K context within 2GB R
Impact-Site-Verification: 9b3ebd10-6c45-44cf-a014-5b8c19b7b285