Topic · 8 stories
8 stories tagged on-device on AI Tamer, newest first.
Opera One's preview downloads a local model, such as Gemma, that sites and extensions can use for text, classification and embeddings, per-site permission.
PrismML's 15 October post adds Ternary Bonsai 2 at 2B, 4B, and 9B. The cards say Apache 2.0 and weights in {-1, 0, +1}.
Liquid AI released LFM2.5-24B-A2B on 13 October, a 24B on-device model it says uses 2.45B active parameters. Commercial use is license-limited.
Google's 7 October post introduces Android on-device AI seal, also called AISeal, a protected virtual machine for personal AI context. It cites SESIP Level 5 and names MediaTek and Snapdragon, with no phone list.
Conway Research opened an invite-only beta of Underdog. TechCrunch says it runs on Mac and Windows now, with Linux and phones still coming. It is free of ads at the start, and revenue is a cut of Stripe purchases.
Google published EmbeddingGemma 2, an open model that maps text, code, images, video, and audio into one 768-wide vector. Developers can load fewer encoders or store a shorter vector.
Magnitude is an Apache-2.0 on-device inference engine for local agents: desktop app plus magnitude CLI, on-hardware kernel compile and tune, shared prefix cache, one-click agents or an OpenAI-compatible API.
Cactus Compute’s Needle 2 (~14MB) turns plain English into local tool calls on a Raspberry Pi 5 CPU—action-only, not a chatbot. RPi News latency demos are attributed to it; the vendor’s ~500 tok/s decode figure is not the same as the table’s ~248–314.
