M5 Max/Ultra users find low-bit DeepSeek beats 27B dense models

Apple M5 Max 128G and M5 Ultra users report that running low-bit quantized DeepSeek 4.1 Flash 1017 locally is far better than BF16 Qwen 3.8 27B, a conclusion endorsed by antirez.

2026-11-11 ~ 2026-11-12 · 2 related posts

• Don't Run BF16 Qwen on M5 Max: Low-Bit DeepSeek Wins on RAM, Speed and Tokens — liuliu · 2026-11-11
• antirez backs low-bit DeepSeek 4.1 Flash over Qwen 27B dense on Apple M5 Macs — antirez · 2026-11-12