Builds / Apple M6 (unified)
Mac mini M6 32 GB: Qwen3.8 Flash-Next (125B-A6B) IQ3_XXS at 11.6 tok/s, 10.4 after a 6K prompt, ~9.4 GB RSS (MoEShift, SSD experts)
Last updated:
- GPU(s)
- Apple M6 (unified)
- VRAM / unified memory (GB)
- 32
- CPU
- Apple M6 (12-core CPU)
- Motherboard
- Mac mini
Amazon · eBay (used) · Newegg - RAM
- 32 GB unified
- PSU
- not stated
- Other
- macOS 27.0.1, Homebrew MoEShift 0.1.0-alpha.1, default config; experts on the internal SSD, GGUF shards on an external SSD; other always-on services left running
- Reported total cost
- not stated
- Power
- not stated
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Prompt processing (tok/s) | Details |
|---|---|---|---|---|---|
| Qwen3.8 Flash-Next | ISTA-DASLab GSQ-RCO IQ3_XXS (~76 GB, 2 shards), experts streamed from the internal SSD | MoEShift 0.1.0-alpha.1 (llama.cpp-based, Metal) | 11.58 | 27.1 @ 72-token prompt | context: 72-token prompt, 512 generated (64K ctx configured)📏 moeshift-community-v1 protocol: median of 3 measured runs after 1 warm-up, 512 output tokens, prompt cache off (cache_n 0); short-prompt runs 11.75 / 11.58 / 11.48; peak combined RSS ~9.4 GB |
| Qwen3.8 Flash-Next | ISTA-DASLab GSQ-RCO IQ3_XXS (~76 GB, 2 shards), experts streamed from the internal SSD | MoEShift 0.1.0-alpha.1 (llama.cpp-based, Metal) | 10.35 | 217 @ 6K-token prompt est. TTFT @8k ≈ 37.8 s (calculated from measured prefill) | context: 6,060-token prompt, 512 generated📏 moeshift-community-v1 protocol: median of 3 measured runs after 1 warm-up, 512 output tokens, prompt cache off (cache_n 0); long-prompt runs 10.34-10.42, prefill 214.7-217.4 |
Prompt processing = prefill speed (how fast the prompt is read) as reported by the source, with the prompt length it was measured at; blank = the source does not report it. est. TTFT @8k is calculated from the measured prefill speed (8,192 ÷ prefill tok/s), shown only when prefill was measured on a prompt of 512+ tokens. It is not a measurement and ignores load time and caching. measured TTFT is the source's own time-to-first-token. Why prefill matters for coding assistants.
Filed 30 Oct 2026, edited 10 Nov (moeshift issue #1). Community report filed on the MoEShift GitHub tracker (MoEShift is an open-source llama.cpp-based engine for Apple Silicon that keeps the shared weights in memory and streams Mixture-of-Experts weights from the SSD as they are needed). It lets the 125B-A6B Qwen3.8 Flash-Next run on Macs with far less memory than the ~76 GB file. Speed depends on SSD speed and expert cache hits, so expect variation. IQ3_XXS is below Q4.
https://github.com/davide-troiani/moeshift
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.