LLaVA on Apple M2 · 16GB

🐢 CPU-only (slow) — LLaVA 13B @ Q4_K_M
LLaVA 13B at Q4_K_M needs ~14.3 GB (weights 7.5 GB + KV 6.3 GB + overhead 512 MB @ 8K ctx) of your 8.2 GB usable (of 16.0 GB unified memory) — runs on CPU only. Expect ~0.5–0.8 tok/s (slow).
context @ Q4_K_M: runs to its full 4K
quantneedsspeed
✗ Q8_020.1 GB—
✗ Q5_K_M15.6 GB—
~ Q4_K_M14.3 GB~0.5–0.8 tok/s (slow)
$runlocal install llava:13b
runlocal verdict card
download card share on x share on reddit drop it in your model-drop post
LLaVA: CPU-only (slow)
for your own README — links back here

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

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

Best models for a Apple M2 · 16GB →

Architecturally similar

These share LLaVA's KV-cache geometry — they size memory the same way, so here's how they fit on a Apple M2 · 16GB too:

modelsharesverdict here
Code Llama 32kv/128hd40kv/128hd 🐢 CPU-only (slow)
OLMo 2 7B 32kv/128hd 🐢 CPU-only (slow)

Check any combo yourself: open the checker →