In 2023 GPT-4 was allegedly 1.8T parameters. In 2026 we have ~100x smaller models (10-20B) that handily outperform it, and can indeed run on a laptop.

▲

rectang 8 minutes ago | parent [-]

How does "outperform" translate to the propensity of an LLM to hallucinate?

	▲	operatingthetan 5 minutes ago \| parent [-]
		There seems to be a mass delusion about how capable SOTA models actually are. That's my only explanation for how poorly I find them performing in basic knowledge tasks compared to how others describe their prowess.