More parameters = more facts stored. Knowledges are almost incompressible, where... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		sinuhe69 2 days ago \| parent \| context \| favorite \| on: DeepSeek v4.1 Flash More parameters = more facts stored. Knowledges are almost incompressible, where strong reasoning only requires a 3B core or so.
		help

yorwba 2 days ago [–]

Weibo's VibeThinker manages with half of that: https://arxiv.org/abs/2511.06221 (They finetuned Qwen2.5-Math-1.5B for reasoning.)

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact