Compact models and agents are bringing inference directly to laptops, phones, and edge devices, reducing cloud dependency and token generation overhead. Recent releases from Liquid AI, Microsoft, and Strawberry highlight a push toward local, responsive, and privacy-aware AI.
A Mac M4 Max test with a Qwen3.8 27B model and 164k tokens of context shows cache and KV can cost as much as weights. The old 64GB rule no longer holds.
Microsoft shows that a 4-billion-parameter vision-language model can hit 66% on OSWorld-Verified, using 13% fewer model calls and 25.4% fewer output tokens after online RL. The secret is a five-stage training pipeline and workflow-first synthetic env...
Liquid AI has introduced LFM3, a family of compact models from 1.7–2.2 billion parameters with a tri-hybrid architecture and llama.cpp support. The decision models skip token generation and return calibrated answers in a single forward pass. For on-p...
An open System One model runs locally on a laptop and reaches results close to GPT-4 on medical benchmarks, with well-controlled calibration on knowledge tests but less reliable patient-level predictions. A two-parameter recalibration method using 50...
A study across seven models and three benchmarks shows that LLM agent skills are not universal: what improves a strong model can degrade a weak one. In 86% of weak-model failures, the violated requirement was already stated in the skill. Adaptation g...
Liquid AI has released a family of models ranging from 1.7 to 2.2 billion parameters, designed to run on phones, laptops, and edge devices. Beyond text, there are multimodal versions and decision models that return answers in a single pass. The commo...
The Swedish startup Strawberry exits beta with an enterprise browser that embeds AI assistants. Agents run on the user's device, not on cloud servers, which can address data sovereignty demands. The company has raised $6 million and plans to scale ac...
