Llama 3.1 on Apple M2 Max · 64GB

✅ Runs (tight) — Llama 3.1 70B @ Q4_K_M
Llama 3.1 70B at Q4_K_M needs ~44.1 GB (weights 39.6 GB + KV 2.5 GB + overhead 2.0 GB @ 8K ctx) of your 45.0 GB usable (of 64.0 GB unified memory) — fits, but little headroom — close other apps or trim context. Expect ~4–7 tok/s (slow).
context @ Q4_K_M: runs to 32K · won't fit past that on this rig
quantneedsspeed
✗ Q8_075.8 GB—
~ Q5_K_M51.3 GB~0.4–0.7 tok/s (slow)
✓ Q4_K_M44.1 GB~4–7 tok/s (slow)
✓ Q3_K_M36.0 GB~5–8 tok/s (slow)
✓ Q2_K28.3 GB~7–11 tok/s (usable)
$runlocal install llama3.1:70b
runlocal verdict card
download card share on x share on reddit drop it in your model-drop post
Llama 3.1: Runs (tight)
for your own README — links back here

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

Also runs on a Apple M2 Max: DeepSeek-R1 (distill) · Qwen3 · Llama 3.3 · Qwen2.5 · Mixtral 8x7B

Best models for a Apple M2 Max · 64GB →

Architecturally similar

These share Llama 3.1's KV-cache geometry — they size memory the same way, so here's how they fit on a Apple M2 Max · 64GB too:

modelsharesverdict here
Qwen2.5-Coder 8kv/128hd ✅ Runs (tight)
Llama 3.2 8kv/128hd ✅ Runs comfortably
Mistral 7B 8kv/128hd ✅ Runs comfortably
Mistral Small 3 8kv/128hd ✅ Runs comfortably
Mistral Nemo 8kv/128hd ✅ Runs comfortably
Codestral 8kv/128hd ✅ Runs comfortably
Llama 3.2 Vision 8kv/128hd ✅ Runs comfortably
Phi-4-mini 8kv/128hd ✅ Runs comfortably

…and 6 more — see the Llama 3.1 page.

Check any combo yourself: open the checker →