Pinned
Insane. We got close to Opus 4.5 at home >70 tokens/s
Replying to @cedric_chee
MiniMax 2.5 full precision FP8 running LOCALLY on vLLM x 8x Pro 6000 🔥
Hosting it is easier then I thought, it just reuse the same script for M2.1. Time to do the vibe coding test!
Generation: 70 tokens-per-sec and 122 tokens-per-sec for two conneciton
Peak Memory: 728GB
