Technology2 min read
What 'Efficient' Actually Means in the Age of Trillion-Parameter Models
Efficiency claims from AI labs rarely specify what's held constant. Here's how to read them.
Nirmay Thakkar
Research
Filter by category, or browse everything we've published.
Efficiency claims from AI labs rarely specify what's held constant. Here's how to read them.
The curve has un-inverted, but strategists disagree on whether it's pricing a soft landing or a delayed one.
Running models on-device is framed as a software win. The real constraint is further upstream.