Test-Time Training via one tool-call
One call to internalize() teaches the model your knowledge through its weights, so it can use what it learns without the source text in context.
GLM 5.3 available now. More models coming.
How learning works
Give your agent something to learn
Send knowledge or memories to internalize(). Here, it’s a company’s return policy.
How serving works
Your agent uses its current memory
Requests use the active version of the model’s learned knowledge.