← sources

HN: I ran 120B MoE models on my 32GB M1 Max using under 16 GB

source: hackernewspublished 2026-11-09 19:20ingested 2026-11-10 02:58

extract

Points: 18 | Comments: 4 Author: tarassenkov Link: https://mitln.github.io/moe-inference-engine/ I ran 120B MoE models on my 32GB M1 Max using under 16 GB

retained text

Points: 36 | Comments: 12
Author: tarassenkov
Link: https://mitln.github.io/moe-inference-engine/
I ran 120B MoE models on my 32GB M1 Max using under 16 GB

document fetched 2026-11-16 02:56