Outlier · Local AI for Mac

Can a Mac run Qwen3.8 27B?

Quick answer

Yes. Qwen3.8 27B has a measured minimum of 24 GB, from build 1.11.792 on 2026-08-19.

We measured 17.2 tokens/s on our Mac Studio M1 Ultra. On the MacBook Pro M1 Pro, expect about 3.4 to 8.4 tokens/s, estimated.

Size and tasks

It can chat, help with code, and read images. Allow 15.5 GB on disk for the download. Its size is 27B dense. By total parameter count, our list includes 19 smaller models and 11 larger models. Dense means the whole model is used for each token.

This Qwen model can help with code, talk through a question and read images. That combination gives you a reason to consider it when you want to keep those jobs in the same download. It has a measured reference speed, so the pace on our test Mac is known. The memory requirement rules out the smallest configurations. Among nearby models, check the supported work as closely as the file size. Sharing a total parameter count does not guarantee that another download can handle the same pictures or coding requests.

The exact model ID is Qwen3.8-27B.

How we know: measured by us. measured 2026-08-19 on M1 Ultra via mlx_vlm 0.6.15 (48-token short-run aggregate); ships in 1.11.792; verified against the shipped sidecar, /health reports 8 tiers including vision38

Which Macs can run it

The memory checks below use the measured minimum.

Check the capacity installed in your Mac against its row below. This model reaches 33 of the 38 Mac models with a suitable memory option. That leaves five Macs without enough memory in any configuration shown. Where a result says tight, the model fits with little room left.

Our M1 Ultra, 64 GB, 800 GB/s test gives the M1 Ultra measurement shown here.

Macs grouped by their smallest fitting memory option
Mac and chipEvery listed memory optionChip speed source
First fitting configuration 24 GB
MacBook Air M2 Apple M2 8 GB does not fit. Next 24 GB configuration.
16 GB does not fit. Next 24 GB configuration.
24 GB tight.
about 1.7 to 5.0 tokens/s estimate, not measured by us
How we estimate speed 1.72 to 5.03 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 100 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro 13 M2 Apple M2 8 GB does not fit. Next 24 GB configuration.
16 GB does not fit. Next 24 GB configuration.
24 GB tight.
about 1.7 to 5.0 tokens/s estimate, not measured by us
How we estimate speed 1.72 to 5.03 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 100 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac mini M2 Apple M2 8 GB does not fit. Next 24 GB configuration.
16 GB does not fit. Next 24 GB configuration.
24 GB tight.
about 1.7 to 5.0 tokens/s estimate, not measured by us
How we estimate speed 1.72 to 5.03 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 100 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Air 13 M3 Apple M3 8 GB does not fit. Next 24 GB configuration.
16 GB does not fit. Next 24 GB configuration.
24 GB tight.
about 1.7 to 4.9 tokens/s estimate, not measured by us
How we estimate speed 1.72 to 4.91 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 100 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Air 15 M3 Apple M3 8 GB does not fit. Next 24 GB configuration.
16 GB does not fit. Next 24 GB configuration.
24 GB tight.
about 1.7 to 4.9 tokens/s estimate, not measured by us
How we estimate speed 1.72 to 4.91 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 100 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
iMac M3 Apple M3 8 GB does not fit. Next 24 GB configuration.
16 GB does not fit. Next 24 GB configuration.
24 GB tight.
about 1.7 to 4.9 tokens/s estimate, not measured by us
How we estimate speed 1.72 to 4.91 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 100 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro 14 M4 Apple M4 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.1 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.06 to 5.54 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 120 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
iMac M4 Apple M4 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.1 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.06 to 5.54 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 120 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac mini M4 Apple M4 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.1 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.06 to 5.54 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 120 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac mini M4 Pro Apple M4 Pro 24 GB tight.
48 GB fits.
64 GB fits.
about 4.7 to 12 tokens/s estimate, not measured by us
How we estimate speed 4.69 to 11.65 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 273 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M4 Pro Apple M4 Pro 24 GB tight.
48 GB fits.
about 4.7 to 12 tokens/s estimate, not measured by us
How we estimate speed 4.69 to 11.65 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 273 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Air 13 M4 Apple M4 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.1 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.06 to 5.54 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 120 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Air 15 M4 Apple M4 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.1 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.06 to 5.54 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 120 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Air 13 M5 Apple M5 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.6 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.63 to 5.51 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 153 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Air 15 M5 Apple M5 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.6 to 5.5 tokens/s estimate, not measured by us
How we estimate speed 2.63 to 5.51 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 153 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro 14 M6 Apple M6 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.6 to 6.2 tokens/s estimate, not measured by us
How we estimate speed 2.63 to 6.18 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 153-170 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M5 Pro Apple M5 Pro 24 GB tight.
48 GB fits.
64 GB fits.
about 5.3 to 11 tokens/s estimate, not measured by us
How we estimate speed 5.28 to 11.46 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 307 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac mini M6 Apple M6 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.6 to 6.2 tokens/s estimate, not measured by us
How we estimate speed 2.63 to 6.18 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 153-170 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac mini M5 Pro Apple M5 Pro 24 GB tight.
48 GB fits.
64 GB fits.
about 5.3 to 11 tokens/s estimate, not measured by us
How we estimate speed 5.28 to 11.46 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 307 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
iMac M6 Apple M6 16 GB does not fit. Next 24 GB configuration.
24 GB tight.
32 GB fits.
about 2.6 to 6.2 tokens/s estimate, not measured by us
How we estimate speed 2.63 to 6.18 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 153-170 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
First fitting configuration 32 GB
MacBook Pro M1 Pro Apple M1 Pro 16 GB does not fit. Next 32 GB configuration.
32 GB fits.
about 3.4 to 8.4 tokens/s estimate, not measured by us
How we estimate speed 3.44 to 8.36 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 200 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M1 Max Apple M1 Max 32 GB fits.
64 GB fits.
about 6.9 to 14 tokens/s estimate, not measured by us
How we estimate speed 6.88 to 14.06 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 400 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M2 Pro Apple M2 Pro 16 GB does not fit. Next 32 GB configuration.
32 GB fits.
about 3.4 to 8.9 tokens/s estimate, not measured by us
How we estimate speed 3.44 to 8.93 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 200 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M2 Max Apple M2 Max 32 GB fits.
64 GB fits.
96 GB fits.
about 6.9 to 15 tokens/s estimate, not measured by us
How we estimate speed 6.88 to 15.14 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 400 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
First fitting configuration 36 GB
MacBook Pro M3 Pro Apple M3 Pro 18 GB does not fit. Next 36 GB configuration.
36 GB fits.
about 2.6 to 7.1 tokens/s estimate, not measured by us
How we estimate speed 2.58 to 7.06 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 150 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M3 Max Apple M3 Max 36 GB fits.
48 GB fits.
64 GB fits.
96 GB fits.
128 GB fits.
about 5.2 to 15 tokens/s estimate, not measured by us
How we estimate speed 5.16 to 15.23 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 300-400 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M4 Max Apple M4 Max 36 GB fits.
48 GB fits.
64 GB fits.
128 GB fits.
about 7.1 to 19 tokens/s estimate, not measured by us
How we estimate speed 7.05 to 19.08 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 410-546 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro M5 Max Apple M5 Max 36 GB fits.
48 GB fits.
64 GB fits.
128 GB fits.
about 7.9 to 21 tokens/s estimate, not measured by us
How we estimate speed 7.91 to 20.73 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 460-614 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac Studio M5 Max Apple M5 Max 36 GB fits.
48 GB fits.
64 GB fits.
128 GB fits.
about 7.9 to 21 tokens/s estimate, not measured by us
How we estimate speed 7.91 to 20.73 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 460-614 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
First fitting configuration 64 GB
Mac Studio M1 Ultra Apple M1 Ultra 64 GB fits.
128 GB fits.
17.2 tokens/s measured on our M1 Ultra
How we estimate speed 17.2 tokens/s measured on an M1 Ultra 64 GB, 800 GB/s
Mac Studio M2 Ultra Apple M2 Ultra 64 GB fits.
128 GB fits.
192 GB fits.
about 14 to 22 tokens/s estimate, not measured by us
How we estimate speed 13.76 to 21.64 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 800 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
First fitting configuration 96 GB
Mac Studio M3 Ultra Apple M3 Ultra 96 GB fits.
256 GB fits.
512 GB fits.
about 14 to 21 tokens/s estimate, not measured by us
How we estimate speed 14.08 to 21.16 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 819 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac Studio M5 Ultra Apple M5 Ultra 96 GB fits.
256 GB fits.
512 GB fits.
about 21 to 31 tokens/s estimate, not measured by us
How we estimate speed 20.64 to 30.96 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 1200 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
No fitting configuration listed
MacBook Air M1 Apple M1 8 GB does not fit. Next 24 GB on a different listed Mac.
16 GB does not fit. Next 24 GB on a different listed Mac.
about 1.2 to 3.3 tokens/s estimate, not measured by us
How we estimate speed 1.17 to 3.26 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 68.25 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Pro 13 M1 Apple M1 8 GB does not fit. Next 24 GB on a different listed Mac.
16 GB does not fit. Next 24 GB on a different listed Mac.
about 1.2 to 3.3 tokens/s estimate, not measured by us
How we estimate speed 1.17 to 3.26 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 68.25 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
Mac mini M1 Apple M1 8 GB does not fit. Next 24 GB on a different listed Mac.
16 GB does not fit. Next 24 GB on a different listed Mac.
about 1.2 to 3.3 tokens/s estimate, not measured by us
How we estimate speed 1.17 to 3.26 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 68.25 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
iMac M1 Apple M1 8 GB does not fit. Next 24 GB on a different listed Mac.
16 GB does not fit. Next 24 GB on a different listed Mac.
about 1.2 to 3.3 tokens/s estimate, not measured by us
How we estimate speed 1.17 to 3.26 tokens/s; estimate: low end from memory bandwidth (Mac bandwidth 68.25 GB/s against 800 GB/s), high end from this chip's speed in llama.cpp's public Apple-silicon benchmark; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s
MacBook Neo Apple A18 Pro 8 GB does not fit. Next 24 GB on a different listed Mac. about 1.0 to 1.4 tokens/s estimate, not measured by us
How we estimate speed 1.03 to 1.42 tokens/s; estimate from memory bandwidth; measured on an M1 Ultra 64 GB: 17.2 tokens/s at 800 GB/s; Mac bandwidth 60 GB/s

Nearby models by total parameter count

Compare these similar-sized downloads if you want a different memory requirement or set of capabilities.

You could use Qwen3.6 27B to chat and help with code. Its measured minimum is the same 24 GB of memory. Its download uses 0.37 GB less disk space. Its size is 27B dense. Switching means giving up image reading.

Gemma 3 27B (QAT) can chat and read images, with a model size of 27B dense. Its download uses 1.33 GB more disk space. Its long-work minimum is 64 GB of memory, 40 GB more. Switching means giving up coding help.

To chat, you could try Gemma 4 26B-A4B. Its measured minimum is 16 GB of memory, 8 GB less. Its size is 26B MoE / 4B active. Its download uses 0.11 GB more disk space. Switching means giving up coding help and image reading.

See the 24 GB memory page to compare the other models you could run with that much memory.

Context and data dates

Memory minimums use a 32K context; longer contexts need more memory. This model starts with a 32768-token context. That is the text it can keep in view as you talk. When planning longer conversations, leave more memory available than the minimum shown here.

The model source date is 2026-08-19. Figures as of 8 November 2026.

Mac source.

Apple technical specifications (apple.com/mac and the specs page of each Mac, read 2026-11-07; the iMac M6 page says only "up to 170GB/s", so its low end, 153 GB/s, is the 16 GB figure for the same M6 chip on the Mac mini page); table last changed 2026-11-07

Chip ratio source.

llama.cpp's public Apple-silicon benchmark (github.com/ggml-org/llama.cpp/discussions/4167, read 2026-11-07): LLaMA 7B Q4_0 text generation, as a ratio to the M1 Ultra 48-core GPU

Model date source.

Dates come from each source text; otherwise table last changed 2026-10-19

Outlier Pro: $39 once

Download free Buy Pro

Free to start (Nano and Lite). macOS 15+, Apple silicon. Refund window: 30 days, no questions asked. Priced in US dollars. Checkout shows your currency and any local tax.

Try Outlier on your Mac

Choose a Mac configuration with enough memory for Qwen3.8 27B, then download Outlier to get started.

Download Outlier