1.2 Trillion Parameters vs. 150 Billion: Google Gemini to Power-Up ‘Linwood’ Siri

It’s a battle of parameters, and Apple has conceded the first round. Google’s “ultrapowerful” 1.2 trillion parameter Gemini model is being integrated into Apple’s ecosystem for a $1 billion annual fee, a monumental leap from Apple’s current 150-billion parameter models.

This integration is part of Apple’s “Glenwood” project, a high-priority effort to fix the notoriously limited Siri. The new assistant, codenamed “Linwood” and set for a spring release, will be a hybrid system, blending Apple’s tech with Google’s.

Google won the coveted spot after an extensive “bake-off,” proving its 1.2T model was superior to OpenAI’s ChatGPT and Anthropic’s Claude. It will be responsible for the complex “summariser” and “planner” functions that the new Siri needs to compete.

The architecture means simple tasks remain on Apple’s servers, but complex, multi-step commands will be routed to the Gemini model. Crucially for privacy, this 1.2T model will run on Apple’s Private Cloud Compute servers, “walled-off” from Google’s infrastructure.

This deal is a “temporary fix” as Apple’s top brass, Craig Federighi and Mike Rockwell, push their own teams to build a 1T+ model. However, catching Google’s rapidly evolving AI is a massive challenge, suggesting this “interim” solution could be semi-permanent.