---
title: "Can I Run Qwen3.8-27B on a Mac? RAM & Speed by Mac"
canonical: https://llmcheck.net/can-i-run/qwen38-27b/
markdown: https://llmcheck.net/md/can-i-run/qwen38-27b.md
page_type: can-i-run
publisher: "LLMCheck (llmcheck.net)"
data_version: 2026-11-18
catalog_verified_through: 2026-11-18
updated: 2026-11-19
license: "CC BY 4.0"
license_url: https://creativecommons.org/licenses/by/4.0/
cite: "LLMCheck. (2026). Can I run Qwen3.8-27B on a Mac? llmcheck.net. https://llmcheck.net/can-i-run/qwen38-27b/ (updated 2026-11-19). CC BY 4.0."
provenance:
  sourced: "Published figure — vendor or press review"
  community: "Linked individual run — blog, forum or submission"
  estimated: "Derived from published bandwidth model (https://llmcheck.net/methodology#estimation)"
---

# Can I run Qwen3.8-27B on a Mac?

Markdown twin of https://llmcheck.net/can-i-run/qwen38-27b/, generated from the LLMCheck dataset (data version 2026-11-18) with the same figures as the page. Every speed names its provenance: **sourced** (a published vendor or press figure), **community** (one person's linked run) or **estimated** (the bandwidth model). LLMCheck is an index and runs no tests of its own. Licence CC BY 4.0: credit "LLMCheck (llmcheck.net)" and link the page.

Yes: **Qwen3.8-27B** runs on 43 of the 54 Macs in the LLMCheck index. It needs about **19 GB** at a 4K context, 24 GB at 64K. Smallest current Mac: **Mac mini M6 24 GB** (raised GPU limit); cheapest at 15+ tok/s: **Mac mini M5 Pro 24 GB,** $1,699 (raised limit). Best measured figure: **55 tok/s** on an M5 Ultra (published).

27.8B, Apache 2.0, by Alibaba; minimum 19 GB of unified memory (weights plus a short context). It loads on 43 of the 54 Mac configurations the index ranks; 32 of them hold it within macOS's default GPU memory limit.

## Verdict for every Mac (54 configurations)

Verdicts: **Runs** = fits macOS's default GPU memory limit (at 15+ tok/s where the index has a figure); **Runs, slowly** = fits the default limit, under 15 tok/s; **Raise GPU limit** = fits the Mac's RAM, but needs more than macOS's default GPU memory limit; the command shown raises the limit until the next restart; **Doesn't fit** = does not load: the model's minimum RAM is above the Mac's, or it has no Mac runtime. Max context = the longest context that fits with a 16-bit KV cache (within the default limit, or the raised one where marked; an estimate from the model's config.json).

| Mac | Sold | Verdict | Build | tok/s | Provenance | Max context | Raise the limit | Model page |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| [MacBook Air M5 16 GB](https://llmcheck.net/best-llm/macbook-air-m5-16gb/) | current | Doesn't fit | — | — | — | — | — | — |
| [MacBook Air M5 24 GB](https://llmcheck.net/best-llm/macbook-air-m5-24gb/) | current | Raise GPU limit | Q4_K_M | 8 | estimated | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | — |
| [MacBook Air M5 32 GB](https://llmcheck.net/best-llm/macbook-air-m5-32gb/) | current | Runs, slowly | Q4_K_M | 8 | estimated | 92K | — | — |
| [MacBook Air M1 8 GB](https://llmcheck.net/best-llm/macbook-air-m1-8gb/) | refurb | Doesn't fit | — | — | — | — | — | — |
| [MacBook Air M4 16 GB](https://llmcheck.net/best-llm/macbook-air-m4-16gb/) | used | Doesn't fit | — | — | — | — | — | — |
| [MacBook Air M2 16 GB](https://llmcheck.net/best-llm/macbook-air-m2-16gb/) | refurb | Doesn't fit | — | — | — | — | — | — |
| [MacBook Air M3 16 GB](https://llmcheck.net/best-llm/macbook-air-m3-16gb/) | refurb | Doesn't fit | — | — | — | — | — | — |
| [MacBook Air M1 16 GB](https://llmcheck.net/best-llm/macbook-air-m1-16gb/) | refurb | Doesn't fit | — | — | — | — | — | — |
| [MacBook Air M4 24 GB](https://llmcheck.net/best-llm/macbook-air-m4-24gb/) | used | Raise GPU limit | Q4_K_M | 7 | estimated | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | — |
| [MacBook Air M2 24 GB](https://llmcheck.net/best-llm/macbook-air-m2-24gb/) | refurb | Raise GPU limit | Q4_K_M | 5 | estimated | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | — |
| [MacBook Air M3 24 GB](https://llmcheck.net/best-llm/macbook-air-m3-24gb/) | refurb | Raise GPU limit | Q4_K_M | 5 | estimated | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | — |
| [MacBook Pro M6 16 GB](https://llmcheck.net/best-llm/macbook-pro-m6-16gb/) | current | Doesn't fit | — | — | — | — | — | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [MacBook Pro M5 Pro 24 GB](https://llmcheck.net/best-llm/macbook-pro-m5-pro-24gb/) | current | Raise GPU limit | 4-bit | 17 | sourced | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | https://llmcheck.net/models/qwen38-27b-on-m5-pro/ |
| [MacBook Pro M6 24 GB](https://llmcheck.net/best-llm/macbook-pro-m6-24gb/) | current | Raise GPU limit | 4-bit | 10 | community | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [MacBook Pro M6 32 GB](https://llmcheck.net/best-llm/macbook-pro-m6-32gb/) | current | Runs, slowly | 4-bit | 10 | community | 92K | — | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [MacBook Pro M5 Pro 48 GB](https://llmcheck.net/best-llm/macbook-pro-m5-pro-48gb/) | current | Runs | 4-bit | 17 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-pro/ |
| [MacBook Pro M5 Max 64 GB](https://llmcheck.net/best-llm/macbook-pro-m5-max-64gb/) | current | Runs | 4-bit | 31 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-max/ |
| [MacBook Pro M5 Pro 64 GB](https://llmcheck.net/best-llm/macbook-pro-m5-pro-64gb/) | current | Runs | 4-bit | 17 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-pro/ |
| [MacBook Pro M5 Max 128 GB](https://llmcheck.net/best-llm/macbook-pro-m5-max-128gb/) | current | Runs | 4-bit | 31 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-max/ |
| [MacBook Pro M3 Max 64 GB](https://llmcheck.net/best-llm/macbook-pro-m3-max-64gb/) | refurb | Runs | Q4_K_M | 19 | estimated | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m3-max/ |
| [MacBook Pro M3 Max 128 GB](https://llmcheck.net/best-llm/macbook-pro-m3-max-128gb/) | refurb | Runs | Q4_K_M | 19 | estimated | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m3-max/ |
| [Mac mini M6 16 GB](https://llmcheck.net/best-llm/mac-mini-m6-16gb/) | current | Doesn't fit | — | — | — | — | — | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [Mac mini M5 Pro 24 GB](https://llmcheck.net/best-llm/mac-mini-m5-pro-24gb/) | current | Raise GPU limit | 4-bit | 17 | sourced | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | https://llmcheck.net/models/qwen38-27b-on-m5-pro/ |
| [Mac mini M6 24 GB](https://llmcheck.net/best-llm/mac-mini-m6-24gb/) | current | Raise GPU limit | 4-bit | 10 | community | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [Mac mini M6 32 GB](https://llmcheck.net/best-llm/mac-mini-m6-32gb/) | current | Runs, slowly | 4-bit | 10 | community | 92K | — | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [Mac mini M5 Pro 48 GB](https://llmcheck.net/best-llm/mac-mini-m5-pro-48gb/) | current | Runs | 4-bit | 17 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-pro/ |
| [Mac mini M5 Pro 64 GB](https://llmcheck.net/best-llm/mac-mini-m5-pro-64gb/) | current | Runs | 4-bit | 17 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-pro/ |
| [Mac mini M2 Pro 16 GB](https://llmcheck.net/best-llm/mac-mini-m2-pro-16gb/) | refurb | Doesn't fit | — | — | — | — | — | — |
| [Mac mini M4 16 GB](https://llmcheck.net/best-llm/mac-mini-m4-16gb/) | used | Doesn't fit | — | — | — | — | — | — |
| [Mac mini M4 Pro 24 GB](https://llmcheck.net/best-llm/mac-mini-m4-pro-24gb/) | used | Raise GPU limit | Q4_K_M | 13 | estimated | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | https://llmcheck.net/models/qwen38-27b-on-m4-pro/ |
| [Mac mini M4 24 GB](https://llmcheck.net/best-llm/mac-mini-m4-24gb/) | used | Raise GPU limit | Q4_K_M | 7 | estimated | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | — |
| [Mac mini M2 Pro 32 GB](https://llmcheck.net/best-llm/mac-mini-m2-pro-32gb/) | refurb | Runs, slowly | Q4_K_M | 11 | estimated | 38K | — | — |
| [Mac mini M4 Pro 48 GB](https://llmcheck.net/best-llm/mac-mini-m4-pro-48gb/) | used | Runs, slowly | Q4_K_M | 13 | estimated | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m4-pro/ |
| [Mac mini M4 Pro 64 GB](https://llmcheck.net/best-llm/mac-mini-m4-pro-64gb/) | used | Runs, slowly | Q4_K_M | 13 | estimated | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m4-pro/ |
| [iMac M6 16 GB](https://llmcheck.net/best-llm/imac-m6-16gb/) | current | Doesn't fit | — | — | — | — | — | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [iMac M6 24 GB](https://llmcheck.net/best-llm/imac-m6-24gb/) | current | Raise GPU limit | 4-bit | 10 | community | 18K with the GPU limit raised | `sudo sysctl iogpu.wired_limit_mb=20480` | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [iMac M6 32 GB](https://llmcheck.net/best-llm/imac-m6-32gb/) | current | Runs, slowly | 4-bit | 10 | community | 92K | — | https://llmcheck.net/models/qwen38-27b-on-m6/ |
| [Mac Studio M5 Max 36 GB](https://llmcheck.net/best-llm/mac-studio-m5-max-36gb/) | current | Runs | Q4_K_M | 24 | estimated | 139K | — | https://llmcheck.net/models/qwen38-27b-on-m5-max/ |
| [Mac Studio M5 Max 48 GB](https://llmcheck.net/best-llm/mac-studio-m5-max-48gb/) | current | Runs | 4-bit | 31 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-max/ |
| [Mac Studio M5 Max 64 GB](https://llmcheck.net/best-llm/mac-studio-m5-max-64gb/) | current | Runs | 4-bit | 31 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-max/ |
| [Mac Studio M5 Ultra 96 GB](https://llmcheck.net/best-llm/mac-studio-m5-ultra-96gb/) | current | Runs | oQ4e | 48 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-ultra/ |
| [Mac Studio M5 Max 128 GB](https://llmcheck.net/best-llm/mac-studio-m5-max-128gb/) | current | Runs | 4-bit | 31 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-max/ |
| [Mac Studio M5 Ultra 256 GB](https://llmcheck.net/best-llm/mac-studio-m5-ultra-256gb/) | current | Runs | oQ4e | 48 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-ultra/ |
| [Mac Studio M5 Ultra 512 GB](https://llmcheck.net/best-llm/mac-studio-m5-ultra-512gb/) | current | Runs | oQ4e | 48 | sourced | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m5-ultra/ |
| [Mac Studio M2 Max 32 GB](https://llmcheck.net/best-llm/mac-studio-m2-max-32gb/) | refurb | Runs | Q4_K_M | 21 | estimated | 38K | — | — |
| [Mac Studio M1 Ultra 64 GB](https://llmcheck.net/best-llm/mac-studio-m1-ultra-64gb/) | refurb | Runs | Q4_K_M | 38 | estimated | 256K (full window) | — | — |
| [Mac Studio M4 Max 64 GB](https://llmcheck.net/best-llm/mac-studio-m4-max-64gb/) | used | Runs | Q4_K_M | 26 | estimated | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m4-max/ |
| [Mac Studio M2 Max 64 GB](https://llmcheck.net/best-llm/mac-studio-m2-max-64gb/) | refurb | Runs | Q4_K_M | 21 | estimated | 256K (full window) | — | — |
| [Mac Studio M1 Ultra 128 GB](https://llmcheck.net/best-llm/mac-studio-m1-ultra-128gb/) | refurb | Runs | Q4_K_M | 38 | estimated | 256K (full window) | — | — |
| [Mac Studio M2 Ultra 128 GB](https://llmcheck.net/best-llm/mac-studio-m2-ultra-128gb/) | refurb | Runs | Q4_K_M | 38 | estimated | 256K (full window) | — | — |
| [Mac Studio M4 Max 128 GB](https://llmcheck.net/best-llm/mac-studio-m4-max-128gb/) | used | Runs | Q4_K_M | 26 | estimated | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m4-max/ |
| [Mac Studio M2 Ultra 192 GB](https://llmcheck.net/best-llm/mac-studio-m2-ultra-192gb/) | refurb | Runs | Q4_K_M | 38 | estimated | 256K (full window) | — | — |
| [Mac Studio M3 Ultra 256 GB](https://llmcheck.net/best-llm/mac-studio-m3-ultra-256gb/) | used | Runs | 4-bit | 38 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m3-ultra/ |
| [Mac Studio M3 Ultra 512 GB](https://llmcheck.net/best-llm/mac-studio-m3-ultra-512gb/) | refurb | Runs | 4-bit | 38 | community | 256K (full window) | — | https://llmcheck.net/models/qwen38-27b-on-m3-ultra/ |

Provenance: **sourced** = Published figure — vendor or press review; **community** = Linked individual run — blog, forum or submission; **estimated** = Derived from published bandwidth model (https://llmcheck.net/methodology#estimation).

For dense models, the formula’s typical error is about **9%** (the median gap to 75 measured plain-decode runs; it tends to undershoot (median −6%), and an estimate scaled from the model’s own measured run is usually closer; [method](https://llmcheck.net/methodology#typical-error)).

## Measured runs of Qwen3.8-27B

| Chip | RAM | Runtime | Quant | tok/s | Provenance | Context | Date | Source | Flags |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| M5 Ultra | 256 GB | LM Studio | Q4_K_M | 55 | sourced | — | 2026-09 | https://www.bgr.com/2264114/mac-studio-m5-ultra-review/ | — |
| M5 Ultra | 256 GB | MLX | 4-bit | 54.7 | community | 512-token prompt, 128 generated | 2026-09-20 | https://github.com/theNetworkChuck/mac-studio-m5-ultra-local-ai | — |
| M5 Ultra | 256 GB | MLX | oQ4e | 48 | sourced | 6,091-token prompt | 2026-09 | https://www.macstories.net/stories/m5-ultra-mac-studio-review-the-dream-mac-for-local-ai-agents/ | — |
| M5 Ultra | 256 GB | llama.cpp | Q4_K_M | 46.27 | sourced | ISL 2048 / OSL 512, depth 0 | 2026-09-21 | https://www.tomshardware.com/desktops/mini-pcs/apple-mac-studio-m5-ultra-review | — |
| M5 Ultra | 512 GB | MLX | BF16 | 19.69 | sourced | 8,192-token context, 128 generated | 2026-10-15 | https://machinelearning.apple.com/research/mlx-distributed-inference | — |
| M3 Ultra | 96 GB | LM Studio | Q4_K_M | 40 | sourced | — | 2026-09 | https://www.bgr.com/2264114/mac-studio-m5-ultra-review/ | — |
| M3 Ultra | 512 GB | MLX | 4-bit | 37.6 | community | 512-token prompt, 128 generated | 2026-09-20 | https://github.com/theNetworkChuck/mac-studio-m5-ultra-local-ai | — |
| M3 Ultra | 512 GB | MLX | oQ4e | 31 | sourced | 6,091-token prompt | 2026-09 | https://www.macstories.net/stories/m5-ultra-mac-studio-review-the-dream-mac-for-local-ai-agents/ | — |
| M3 Ultra | 512 GB | MLX | BF16 | 12.46 | sourced | 8,192-token context, 128 generated | 2026-10-15 | https://machinelearning.apple.com/research/mlx-distributed-inference | — |
| M5 Max | 128 GB | LM Studio | 4-bit | 32.9 | community | 512-token coding answer after a short prompt, median of 5 at T 0.6 | 2026-10-25 | https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-25-m5-max/README.md | — |
| M5 Max | 128 GB | MLX | 4-bit | 31 | community | 512-token coding answer after a short prompt, median of 5 at T 0.6 | 2026-10-25 | https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-25-m5-max/README.md | — |
| M5 Max | 128 GB | MLX | 8-bit | 17.9 | community | short prompt, 400 tokens | 2026-08-20 | https://github.com/nicedreamzapp/claude-code-local/blob/main/docs/BENCHMARKS.md | — |
| M1 Max | 64 GB | LM Studio | 8-bit | 11.38 | community | 41-token chat prompt, 541 generated, thinking on (medium) | 2026-10-08 | https://github.com/Humanoid-Intel/llmcheck-data/issues/1 | — |
| M5 Pro | 48 GB | LM Studio | 4-bit | 16.81 | community | 512-token coding answers, median of 5 at T 0.6 | 2026-10-19 | https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-19-m5pro-48gb/README.md | — |
| M5 Pro | 48 GB | llama.cpp | 4-bit | 16.5 | sourced | single stream; prompt length not stated | 2026-10-15 | https://www.frandroid.com/test/3283093_test-nvidia-rtx-spark | — |
| M5 Pro | 48 GB | MLX | 4-bit | 16.33 | community | 512-token coding answers, median of 5 at T 0.6 | 2026-10-19 | https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-19-m5pro-48gb/README.md | — |
| M6 | 32 GB | MLX | DWQ | 11.5 | community | ~20-token chat prompt, up to 400 generated | 2026-10-28 | https://github.com/suhasaraos/mac-mini-local-llm | — |
| M6 | 32 GB | MLX | MXFP4 | 10.2 | community | ~20-token chat prompt, up to 400 generated | 2026-10-28 | https://github.com/suhasaraos/mac-mini-local-llm | — |
| M6 | 24 GB | MLX | 4-bit | 9.8 | community | 525-token prompt | 2026-10-09 | https://llamaperf.com/report/a531403b-f06b-4efe-b033-c067a3c68b0f | not an anchor |
| M6 | 32 GB | MLX | 4-bit | 9.8 | community | 1k | 2026-11-05 | https://omlx.ai/benchmarks/performance/6tn94hxd | — |
| M6 | 32 GB | MLX | 4-bit | 9.7 | community | ~20-token chat prompt, up to 400 generated | 2026-10-28 | https://github.com/suhasaraos/mac-mini-local-llm | — |
| M6 | 24 GB | LM Studio | QAT | 9.42 | community | short prompt | 2026-09-22 | https://nyaa.sh/reviews/mac-mini-m6-local-ai | — |

One row is one measurement session; on each chip the verdict table shows the lower median of the eligible runs (plain decode at 5 bits or fewer). A binned or not-an-anchor run is shown but never used as a reference.

## Run it

Ollama tag, verified on ollama.com 2026-11-14: `ollama run qwen3.8:27b`

Model card: https://huggingface.co/Qwen/Qwen3.8-27B


## Cite this page

LLMCheck. (2026). Can I run Qwen3.8-27B on a Mac? llmcheck.net. https://llmcheck.net/can-i-run/qwen38-27b/ (updated 2026-11-19). CC BY 4.0.

How to cite: https://llmcheck.net/cite · Open data (CSV, JSON): https://llmcheck.net/data/ · Every Markdown mirror: https://llmcheck.net/md/index.md
