The New Open-Weights Leader, AI Lobbying, On-Device Reasoning, and Sleep-Based Disease Prediction

GLM-5 Scales Up

TLDR: Z.ai released GLM-5, a 744 billion-parameter mixture-of-experts model that now leads the open-weights leaderboard on Artificial Analysis' Intelligence Index. It nearly matches proprietary models like Claude Opus 4.6 and GPT-5.2 while remaining open under an MIT license.

Link: GLM-5 on Artificial Analysis

Big AI Spends Big on Lobbying

TLDR: Top tech and AI companies spent more than $100 million on political lobbying in 2025, with Meta leading at $26.29 million. The spending appears to be paying off through favorable policy outcomes including relaxed chip export bans, limits on state AI regulation, and support for massive infrastructure projects.

Link: Bloomberg report on tech lobbying

Faster Reasoning at the Edge

TLDR: Liquid AI released LFM2.5-1.2B-Thinking, a reasoning model that runs in under 900 megabytes of RAM and generates tokens roughly twice as fast as similarly sized competitors. It matches or beats Qwen3-1.7B on most reasoning benchmarks while requiring significantly less memory.

Link: Liquid AI LFM2.5-1.2B-Thinking

Sleep Signals Predict Illness

TLDR: SleepFM, developed by researchers at Stanford, Harvard, and several European institutions, can classify over 130 medical conditions from a single night of sleep data, detecting conditions like congestive heart failure and stroke up to six years before symptoms appear.

Link: SleepFM on GitHub