Yes, just — 1.1 GB spare
LLM-jp 4 33B Thinking at Q4_K_M fits your L4 entirely on the GPU at 8K context, at an estimated 9.7 tokens per second. Past 12K the KV cache pushes it over — quantise the cache to q8_0, or step down a quantisation, to go longer.
Fully on GPU
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
Free 1.1 GB of 22.4 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 32.9 GB | 35.5 GB | — | ~2.1 | −0.1% ppl | 13.1 GB over |
| Q6_K | 25.4 GB | 28.0 GB | — | ~3.8 | −0.4% ppl | 5.6 GB over |
| Q5_K_M | 21.9 GB | 24.5 GB | — | ~6.0 | −0.8% ppl | 2.1 GB over |
| Q4_K_M | 18.7 GB | 21.3 GB | 12K | 9.7 | −1.9% ppl | Recommended |
| Q3_K_M | 15.1 GB | 17.7 GB | 26K | 12 | −5.4% ppl | Long context |
| Q2_K | 12.9 GB | 15.5 GB | 35K | 14 | −15% ppl | Long context |
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
$ llama-server \
-hf llm-jp/llm-jp-4-33b-thinking:Q4_K_M \
-c 8192 -ngl 99
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.
03Only 1.1 GB is spare, so a long system prompt can still push it over. Its real ceiling here is 12K context.