More
Submit build

Builds / Apple M1 Pro 14-core GPU (unified)

MacBook Pro M1 Pro 32 GB (2021): Qwen3.8 Flash-Next (125B-A6B) IQ3_XXS at 7.0 tok/s, 5.3 after a 6K prompt (MoEShift, SSD experts)

Last updated:

Source: github.com ↗More source: github.com ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
GPU(s)
Apple M1 Pro 14-core GPU (unified)
VRAM / unified memory (GB)
32
CPU
Apple M1 Pro (10-core CPU)
Motherboard
MacBook Pro 14-inch (2021)
Amazon · eBay (used) · Newegg
RAM
32 GB unified
PSU
not stated
Other
macOS 27.2.1, MoEShift 0.1.0-alpha.5 from source (a9f103c), all defaults (tuned for M4 Pro 24 GB): ctx 65536, q8_0 KV, flash attention, 24 protected slots; AC power, light desktop apps open
Reported total cost
not stated
Power
not stated

Measured speed (generation tokens/sec)

ModelQuantEnginetok/sPrompt processing (tok/s)Details
Qwen3.8 Flash-NextISTA-DASLab GSQ-RCO IQ3_XXS (~76 GB, 2 shards), experts streamed from the internal SSDMoEShift 0.1.0-alpha.5 (llama.cpp-based, Metal)7.0217.7 @ 72-token promptcontext: 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 7.07 / 6.72 / 7.02
Qwen3.8 Flash-NextISTA-DASLab GSQ-RCO IQ3_XXS (~76 GB, 2 shards), experts streamed from the internal SSDMoEShift 0.1.0-alpha.5 (llama.cpp-based, Metal)5.33105 @ 6K-token prompt
est. TTFT @8k ≈ 77.8 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 5.33 / 5.51 / 5.28; reporter notes 200-350 MiB of expert reads per token at 45-55% cache-miss rate

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 15 Nov 2026 (moeshift issue #13), first report on an M1-generation Mac. Decode after the 6K prompt is ~24% slower than short-prompt decode here, vs ~5% on M4 Pro. 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.

Source: https://github.com/davide-troiani/moeshift/issues/13
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.

Embed this result

Copy into a README or blog. Links back to this build. See badge docs (CC BY 4.0 attribution). The SVG does not invent speeds — numbers stay on the build page.

Markdown

[![Measured on RigBase](https://getrigbase.com/badge/macbook-pro-m1-pro-32gb-qwen38-flash-next-iq3xxs-moeshift-ssd-experts.svg)](https://getrigbase.com/builds/macbook-pro-m1-pro-32gb-qwen38-flash-next-iq3xxs-moeshift-ssd-experts/)

HTML

<a href="https://getrigbase.com/builds/macbook-pro-m1-pro-32gb-qwen38-flash-next-iq3xxs-moeshift-ssd-experts/" rel="noopener"><img src="https://getrigbase.com/badge/macbook-pro-m1-pro-32gb-qwen38-flash-next-iq3xxs-moeshift-ssd-experts.svg" alt="Measured on RigBase" height="20"/></a>

Preview: Measured on RigBase · How to use

Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)