Mac mini M6 32GB tested: local Qwen3 27B inference hits 22-50 t/s
Striking_Luck_886 · reddit · 2026-10-14
A Reddit user shares first-day benchmarks of the Mac mini M6 32GB, comparing incoai's Splash and trymirari's Uzu local inference stacks—results were similar, and they picked Splash for out-of-the-box uncensored support. Mirai Qwen3.8-27B-M with reasoning disabled: 21.9 t/s on prose, 41.0 t/s on Python, 50.2 t/s on repetitive JSON, and 16.81s first-token latency on a 6,241-token prompt. Real-world speed is roughly 22-50 t/s; speculative decoding helps predictable code and JSON far more than prose.
More from Infra
- ECDSA.fail quantum challenge hits 893 qubits x 2.2M Toffolis, 25% cheaper than Google's best — jedisct1 · 2026-10-14
- Llamaforge v0.16: a control panel for upstream llama.cpp with local agents whose quotes are code-verified — Sleepybear2611 · 2026-10-14
- Questioning OpenAI's Decisions API: are cloud round-trip latency and privacy viable in production? — WhichCardiologist800 · 2026-10-14
- Kev-4B Decision Model Goes Live in Beta on Phala with NVIDIA Confidential Computing — bgmshana · 2026-10-14
- Zuckerberg: Meta committed hundreds of billions on AI infra before knowing it would work — GaryMarcus · 2026-10-14
- NVIDIA's 400G/lane optical spec puts tougher laser demands on CPO — jwt0625 · 2026-10-14