Editing Openai/6939b23e-87cc-8001-bdd9-9546d1143ce9 (section)

==== - Larger models have: - More parameters → can store more factual knowledge internally - Broader training data → more likely to have encountered references to “Samurai Jack” and its characters ====
* When asked, they can recall specifics accurately without needing external context.
* Important distinction: - It’s not the vocabulary size in tokens; it’s how much information is encoded in the model’s weights. - That’s why you can ask ChatGPT about obscure topics it has “memorized,” while a small model fills in gaps with hallucinations.