Qwen3 Coder 30B-A3B on Apple M2 Pro · 32GB

🐢 CPU-only (slow) — Qwen3 Coder 30B-A3B @ Q6_K
Qwen3 Coder 30B-A3B 30B-A3B (MoE) at Q6_K needs ~25.3 GB (weights 23.4 GB + KV 768 MB + overhead 1.2 GB @ 8K ctx) of your 18.4 GB usable (of 32.0 GB unified memory) — runs on CPU only. Expect ~4–6 tok/s (slow).
context @ Q6_K: runs to 32K · won't fit past that on this rig
quantneedsspeed
✗ Q8_032.5 GB—
~ Q6_K25.3 GB~4–6 tok/s (slow)
~ Q4_K_M19.0 GB~5–8 tok/s (slow)
$runlocal install qwen3-coder:30b
runlocal verdict card
download card share on x share on reddit drop it in your model-drop post
Qwen3 Coder 30B-A3B: CPU-only (slow)
for your own README — links back here

Same model, other machines: Apple M1 · Apple M1 Pro · Apple M2 · Apple M3 Pro · Apple M4 Pro · Apple M3 Max

Also runs on a Apple M2 Pro: DeepSeek-R1 (distill) · Mistral Small 3 · Qwen3 · Llama 3.1 · Qwen2.5

Best models for a Apple M2 Pro · 32GB →

Architecturally similar

These share Qwen3 Coder 30B-A3B's KV-cache geometry — they size memory the same way, so here's how they fit on a Apple M2 Pro · 32GB too:

modelsharesverdict here
Qwen2.5-Coder 4kv/128hd ✅ Runs (tight)
StarCoder2 4kv/128hd ✅ Runs (tight)
Yi 1.5 4kv/128hd ✅ Runs (tight)
SmolLM3 3B 4kv/128hd ✅ Runs comfortably
Yi-Coder 9B 4kv/128hd ✅ Runs comfortably

Check any combo yourself: open the checker →