The Sequence AI of the Week #899: Inside Inkling: A Trillion-Parameter Model That Only Wakes Up 41 Billion at a Time
Thinking Machine's new model Inkling has 975 billion parameters but activates only 41 billion per token, using a router to select specialist subsets for sparse computation.
Inkling is best understood not as a single 975-billion-parameter brain that fires all at once, but as a giant warehouse of specialist capacity. A router chooses a small working set for each token. That makes the arithmetic sparse, while storage, networking, and deployment remain very large.
The headline number for Inkling is 975 billion parameters. That is close enough to a trillion that the distinction is mostly useful to accountants. The more interesting number is the one beside it: 41 billion parameters active per token. In other words, the model owns an enormous amount of capacity, but a single word passing through the network touches only about 4.2 percent of it.
A good mental picture is a university with 256 specialist departments on each relevant floor. When a token arrives, a dispatcher does not convene the entire university. It selects six departments that appear useful for this token, adds two general-purpose departments that always attend, combines their work, and moves on. A line of Python might summon one set of specialists; a phrase in Greek, a diagram label, or a piece of audio may summon another.
Read more