Running an M5 Max MacBook in low-power mode full time to trade tok/s for fan-less silence
joshwhiton · x · 2026-09-19
The author runs an M5 Max MacBook in low-power mode all the time, trading slower tok/s for fan-less silence, and notes 128GB unified memory is still plenty for running Qwen locally.
More from Infra
- Agents are eating CPUs: Intel CEO says he can fill only half of orders — demian_ai · 2026-09-19
- Qwen 27B IQ3_XXS beats Bonsai ternary PQ2 by 3x on 16GB VRAM local test — Danmoreng · 2026-09-19
- Cloud engineer debunks $30-60/mo per-user cost claims for Meta's Muse agents via Firecracker microVMs — carlosdponx · 2026-09-19
- mitsuhiko: Renting 1,000 GPUs won't cut it unless you build your own data center — mitsuhiko · 2026-09-19
- Dev calls out inference providers claiming 99% cache rates as likely DeepSeek wrappers — TheZachMueller · 2026-09-19
- Why DeepSeek's Cache Rates Are So Hard to Replicate, Explained — thdxr · 2026-09-19