Only with CPU offload

LLM-jp 4 33B Thinking at Q4_K_M needs 21.3 GB but only 7.0 GB is addressable, so about 76% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 2.4 tokens per second — usable for batch work, painful for chat.

24% on GPU 8K context Q4_K_M · 18.7 GB Apache 2.0 Released 14 Aug 2026 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.

Hardware requirements What hardware do I need to run LLM-jp 4 33B Thinking? →

At Q4_K_M with an 8K context: a 32 GB GPU (RTX 5090) or a 36 GB Mac (M4 Max). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.

The VRAM budget

weights 18.7 GB
Weights 18.7 GB KV cache @ 8K 2.00 GB Runtime overhead 0.6 GB Over budget 14.3 GB past 7.0 GB

Quantisation ladder

QuantWeightsTotal @ 8KMax contextTok/sQualityFit
Q8_0 32.9 GB 35.5 GB — ~1.2 −0.1% ppl 28.5 GB over
Q6_K 25.4 GB 28.0 GB — ~1.7 −0.4% ppl 21.0 GB over
Q5_K_M 21.9 GB 24.5 GB — ~2.0 −0.8% ppl 17.5 GB over
Q4_K_M 18.7 GB 21.3 GB — ~2.4 −1.9% ppl 14.3 GB over
Q3_K_M 15.1 GB 17.7 GB — ~3.1 −5.4% ppl 10.7 GB over
Q2_K 12.9 GB 15.5 GB — ~3.8 −15% ppl 8.5 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

terminal
$ llama-server \
    -hf llm-jp/llm-jp-4-33b-thinking:Q4_K_M \
    -c 8192 -ngl 15

The engine underneath most of the others. Every knob is exposed. More on llama.cpp.

01Download is 18.7 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.
See all models for this rig Compare with another model

Related models on a Radeon RX 6600 XT

ModelTotalTok/s est.
Qwen3 32B
32.8B · similar size
Q4_K_M · 21.0 GB ~2.4 CPU offload
DeepSeek-R1-Distill-Qwen 32B
32.8B · similar size
Q4_K_M · 21.0 GB ~2.4 CPU offload
Qwen2.5-Coder 32B
32.8B · similar size
Q4_K_M · 21.0 GB ~2.4 CPU offload
Olmo 3.1 32B Instruct
32.2B · similar size
Q4_K_M · 20.0 GB ~2.6 CPU offload
Nemotron 3.5 Lightning 30B-A3B
31.6B MoE · similar size
Q4_K_M · 18.4 GB ~11 CPU offload
Gemma 4 31B
31.3B · similar size
Q4_K_M · 20.2 GB ~2.5 CPU offload
GLM-4.7-Flash 30B-A3B
31.2B MoE · similar size
Q4_K_M · 18.6 GB ~11 CPU offload

Every model against the Radeon RX 6600 XT →

LLM-jp 4 33B Thinking on other hardware

DeviceTotalTok/s est.
GeForce RTX 4060 Laptop
7.0 GB usable · 256 GB/s
21.3 GB ~2.4 CPU offload
GeForce RTX 3060 12 GB
10.6 GB usable · 360 GB/s
21.3 GB ~3.0 CPU offload
M4 · 16 GB
10.7 GB usable · 120 GB/s
21.3 GB — Won't fit
GeForce RTX 4090
22.4 GB usable · 1008 GB/s
21.3 GB 33 Tight fit
Radeon RX 7900 XTX
22.4 GB usable · 960 GB/s
21.3 GB 31 Tight fit
GeForce RTX 5090
30.4 GB usable · 1792 GB/s
21.3 GB 58 Runs great
M4 Max · 36 GB
27.0 GB usable · 410 GB/s
21.3 GB 12 Runs great
M3 Max · 64 GB
48.0 GB usable · 400 GB/s
21.3 GB 12 Runs great

All 114 devices, by tier →