
Model deprecation is a shutdown announced in advance. Here's the model lifecycle, the notice periods, and why a silent version swap hurts more than an error.

On June 13 OpenRouter shipped Fusion: one request fans out to a panel of models, and a judge merges them into a single answer while flagging where they disagree. That's a weekend bot for the questions that matter — with an honest 'the models don't agree here' note.

A small model on your laptop answers almost like a giant. Distillation is when the big model personally tutors the small one.

Fine-tuning teaches a model style and behavior, not new facts. Here's how it differs from RAG and why a beginner should care about the difference.

Open weights are a model's ready-made 'brain' you can download and run. But it's not open source: you don't get the recipe or the data. Here's the difference.