Perfect timing! Wanted to self-host a beefy OSS model to add to our Khala model orchestrator; the full GLM 5.2 was too big for the capacity we have on gcloud but this REAP fits on 4x RTX Pros which…