Knowledge Cutoff
Batas Pengetahuan
The date after which the model has no trained knowledge at all.
A model's knowledge cutoff is the date of the most recent data in its training corpus. Everything after that date is unknown to it unless it is provided in the context, fetched by a tool, or the model has been updated since.
The cutoff is the reason models confidently describe last year's prices, quote regulations that have since changed, and have no idea what happened this week. It also explains why newer models with a later cutoff tend to be better at current events even when the benchmark scores are close.
It is a hard boundary, not a soft one. A model does not become vague as it approaches the cutoff; it simply has nothing, and the way it handles that absence depends on whether the surrounding system gives it a way to look things up.
Related terms
Hallucination
Output that is fluent and wrong. Two different problems hide under one word.
RAG
Fetch the relevant documents first, then let the model answer using them.
Grounding
Tying the answer to something checkable, so it can be verified instead of trusted.
Model
The trained file of learned parameters that turns input into output.
Web Search
A tool that lets the model look things up instead of guessing.
Want this working inside your team?
It starts with one conversation about your workflow.