qwen3.6-27b
Run Qwen3.6 27B on a Mac
The dense flagship of the Qwen3.6 line. Dense means every parameter works on every token — highest quality-per-token in this lineup, but dense decode is slow next to MoE models of a similar size. On 32 GB Macs, most people are happier with the 35B MoE, which activates only ~3 B parameters per token.
27 B dense · 4-bitquality-first tasks where speed doesn't matter
Alibaba 256K context Base model intelligence 37.7 AA
One command
rapid-mlx serve qwen3.6-27b
Weights (13.5 GB) download on first run; you get an OpenAI-compatible
endpoint at http://localhost:8000/v1. No rapid-mlx yet? It's one line —
curl -fsSL https://rapidmlx.com/install.sh | bash — or take the
desktop app.
Point a client at it. OpenAI SDK: base_url="http://localhost:8000/v1",
model="qwen3.6-27b". Claude Code:
ANTHROPIC_BASE_URL=http://localhost:8000 (no /v1 suffix) —
see the Claude Code guide.
Measured on real hardware
Headline machine · Mac mini M4 Pro · 48 GB
All measured machines
| Machine | Decode | First token | Peak memory | Cold boot | Weights |
|---|---|---|---|---|---|
| Mac mini M4 Pro · 48 GB rapid-mlx 0.15.5 · 2026-10-03 raw data | 23.4 tok/s | 18.9 s | 22.8 GB | — | 13.5 GB |
Each row cites its own run; numbers are never copied across machines. Mac mini M4 Pro · 48 GB: rapid-mlx 0.15.5, 2026-10-03 — raw data + method — 4 concurrent streams: 25.1 tok/s aggregate.
Community runs, one page per Mac: Apple M4 Max.
Will it fit your Mac?
Measured 22.8 GB Metal peak (15.3 GB process RSS) on a 48 GB M4 Pro across the whole benchmark run — up to a ~16K-token prompt and four concurrent streams, including the engine's prefix cache, which defaults to 20% of RAM. A short chat needs less; long contexts grow the KV cache, so leave headroom.
For the conservative install-default placement see the hardware tiers table; to compare against every model your RAM can hold, use the live picker.
Variants & profiles
qwen3.6-35b— Qwen3.6 35B MoE — much faster decode, ~3 B active per token
FAQ
How much memory does Qwen3.6 27B need on a Mac?
Measured 22.8 GB Metal peak (15.3 GB process RSS) on a 48 GB M4 Pro across the whole benchmark run — up to a ~16K-token prompt and four concurrent streams, including the engine's prefix cache, which defaults to 20% of RAM. A short chat needs less; long contexts grow the KV cache, so leave headroom.
How fast is Qwen3.6 27B on Apple Silicon?
We measured 23.4 tok/s on a Mac mini M4 Pro · 48 GB, rapid-mlx 0.15.5, 2026-10-03; peak memory 22.8 GB on that run. Every row on this page keeps its own chip, version and date — no number is copied across machines.
How do I run Qwen3.6 27B locally?
rapid-mlx serve qwen3.6-27b — the weights download on first run and you get an OpenAI-compatible endpoint at localhost:8000/v1 for Aider, Codex and any OpenAI client, plus the Anthropic API at localhost:8000 for Claude Code.