Don't Run BF16 Qwen on M5 Max: Low-Bit DeepSeek Wins on RAM, Speed and Tokens
liuliu · reddit · 2026-11-11
Reddit user liuliu argues that M5 Max 128G / M5 Ultra owners will be much happier running low-bit DeepSeek 4.1 Flash 1017 than BF16 Qwen 3.8 27B:
- Less RAM: with SSD streaming, can cap at 50G while keeping 30 tok/s
- 5x fewer reasoning tokens than Qwen 3.8 27B
- Near-cloud feel: 4000 tok/s prefill, 100 tok/s decode on M5 Ultra
- No long-context degradation: smartness and performance hold steady at 200k–400k context
The author invites counterarguments.
Related event: M5 Max/Ultra users find low-bit DeepSeek beats 27B dense models(2 posts)→
More from Infra
- Dev builds all-rust AI OS LexiSuite, claims faster Qwen3 27B inference than Ubuntu — W61k3r · 2026-11-12
- KubeCon keynote: computing is shifting from human speed to agent speed — HowDevelop · 2026-11-12
- Leaked memo: Meta infrastructure head Santosh Janardhan to leave in early 2027 — dinabass · 2026-11-12
- Huawei open-sources MACaQ quantizer: 23% lower KL than Unsloth GGUFs — acluk90 · 2026-11-12
- Omarchy 4.6 lets you share local AI models across your tailnet, says DHH — alexcovo_eth · 2026-11-12
- Eric Schmidt's foundation funds French NGO fighting against data centers — StephanSturges · 2026-11-12