Running Jev-style models on a 24GB laptop for three-question decisions

Sudo su @sudoingXOriginally posted 2026.10.06

Open original on X

i pulled both cloudflare's jev like models at q4, clef 27b at 19.2 gb and its 9b sibling at 6.5 gb, onto my rtx 5090 laptop with 24gb of vram 🤯 and the 9b already answers a three question decision in 70 ms, i'm genuinely excited about this one, next i'm letting it play the breakout my bonsai 2 agents built on a 16gb card

This post quotes another post: x.com/cloudflare/status/2105747536510099540