Builds / Apple M5 Pro (unified)
MacBook Pro M5 Pro 24 GB: Qwen3.8 Flash-Next (125B-A6B) IQ3_XXS at 14.7 tok/s with experts streamed from SSD (MoEShift)
Last updated:
- GPU(s)
- Apple M5 Pro (unified)
- VRAM / unified memory (GB)
- 24
- CPU
- Apple M5 Pro
- Motherboard
- MacBook Pro
Amazon · eBay (used) · Newegg - RAM
- 24 GB unified
- PSU
- not stated
- Other
- macOS 26.7.1, MoEShift v0.1.0-alpha.3, default config; normal desktop use during the run (Safari, Teams, IntelliJ open); on mains power
- 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.3 (llama.cpp-based, Metal) | 14.69 | 39.2 @ 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 generation range 13.95-14.87; peak process footprint ~8.5 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.3 (llama.cpp-based, Metal) | 13.14 | 374 @ 6K-token prompt est. TTFT @8k ≈ 21.9 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 generation range 11.57-13.55, prefill 359.9-375.7 |
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 10 Nov 2026 (moeshift issue #4). 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.