No — not on this device
LLM-jp 4 33B Thinking at Q4_K_M needs 21.3 GB against 10.7 GB usable, and the shortfall of 10.6 GB is more than 32 GB of system RAM can cover at a tolerable speed. A smaller sibling or a lower quantisation is the honest answer here.
Does not fit
8K context
Q4_K_M · 18.7 GB
Apache 2.0
Released 14 Aug 2026
New this week
Not in the Ollama library
Japan’s national-institute reasoning model, Japanese and English. A plain dense Llama-style 33B: Q4 is a tight 24 GB fit.
The VRAM budget
weights 18.7 GB
Weights 18.7 GB
KV cache @ 8K 2.00 GB
Runtime overhead 0.6 GB
Over budget 10.6 GB past 10.7 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 32.9 GB | 35.5 GB | — | ~1.2 | −0.1% ppl | 24.8 GB over |
| Q6_K | 25.4 GB | 28.0 GB | — | ~1.5 | −0.4% ppl | 17.3 GB over |
| Q5_K_M | 21.9 GB | 24.5 GB | — | ~1.7 | −0.8% ppl | 13.8 GB over |
| Q4_K_M | 18.7 GB | 21.3 GB | — | ~2.0 | −1.9% ppl | 10.6 GB over |
| Q3_K_M | 15.1 GB | 17.7 GB | — | ~2.5 | −5.4% ppl | 7.0 GB over |
| Q2_K | 12.9 GB | 15.5 GB | — | ~2.9 | −15% ppl | 4.8 GB over |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it.
How to run it
$ pip install mlx-lm $ mlx_lm.generate --model mlx-community/llm-jp-4-33b-thinking-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.
01Download is 18.7 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02macOS caps what the GPU may wire down at about 10.7 GB of 16 GB. Raising it with iogpu.wired_limit_mb is possible, and risky.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.