Google’s Gemma 4 AI models get 3x speed boost by predicting future tokens

The problem with rolling your own AI is that your system memory probably isn’t very fast compared to the high bandwidth memory (HBM) used in enterprise hardware. As a result, the processor spends a lot of time moving parameters from VRAM to compute units for each token, and compute cycles are going unused during this process.
Android Central on MSN

Google's AI search is finally doing a better job of linking to the web

AI Overviews now highlight articles, forums, and subscription content more clearly.

Google appears to have an 'AI Ultra Lite' subscription tier in the works

Google is quietly developing AI Ultra Lite as a new subscription option for Gemini. The price for AI Ultra Lite, codenamed “Neon”, may fall somewhere between the $20 and $250 range offered by AI Pro and AI Ultra. Google is also working on adding a new AI usage limit dashboard.