Running open-weight models through ShadowBench on Modal, with a Plato's cave twist

AAAzzam · x · 2026-11-19

ajhinh got curious how open-weight models fare on ShadowBench and ran a few of them on Modal to find out. AAAzzam then asks how K3.1 does on PlatoBench, with a Plato's cave allegory illustration attached.

A light experimental post probing model behavior on benchmarks; concrete scores are in the linked thread.

Original post →

More from Models

Models channel →