ParentFull threadbavell·IDK about "efficiently" but we've been able to run llms locally with AMD for 1.5-2 years nowView on HN