Ox Alpha was GLM 5.3 Flash, and that's wild

hey have u guys been following the ox alpha mystery on openrouter

some random model showed up with no name, no company, no logo. just listed as stealth with 1M context and free access. devs went crazy trying to figure out who made it

turns out it was z.ai's GLM 5.3 flash. 320B params, only 18B active per token, MIT licensed, open weights on huggingface now. and they ran the whole thing on like 100k chinese made chips with no nvidia hardware at all

been testing it and honestly the quality is solid. $0.15 in and $0.50 out per million tokens. that's insanely cheap for what you get

the anonymous launch was kinda genius tbh. no branding bias, just pure quality. it topped the charts before anyone knew who made it

anyone deployed it in prod yet? curious about real world latency vs the benchmark claims

harshkadyan #thedevs #brutalharsh