Why Frontier AI Models Fail at Biology | Alexander Amini

Specialized, locally-run foundation models can outperform frontier cloud LLMs on narrow domains like biological aging prediction and robotics.

• LFM 2.5 24B MoE launched: Liquid AI's new model has 24B total parameters with only 2B active, runs at 2B-model speed, fits in 16GB RAM for AI PCs, and targets tool-calling/agentic tasks.
• Specialists beat generalists: Amini argues frontier cloud models (ChatGPT, Claude, Gemini) are broad generalists but poor specialists; narrow-trained local models can exceed frontier capability on specific tasks.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.