Knowledge cutoff
The date after which a model simply hasn’t seen anything — its news stops there.
Think of it like
Like a textbook printed last year: thorough up to press time, blank on everything since.
Example
Asked who won a championship held after its cutoff, the model either admits it doesn’t know or confidently guesses wrong.
How it actually works
A model’s parametric knowledge is frozen at the end of its training data. Anything later is invisible unless you supply it in the prompt or via retrieval. The subtle danger is that models don’t reliably know what they don’t know, so they can answer post-cutoff questions with unwarranted confidence.
For product teams
Why current-events features need live retrieval bolted on, not just a bigger model.
For engineers
The training-data recency bound of parametric knowledge; post-cutoff facts require in-context grounding.
Related
- Parametric knowledge — The frozen knowledge it bounds.
- RAG — How you get around it.
- Hallucination — The confident-wrong risk it creates.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome