Yes — with 196.6 GB to spare
Gemma 4 12B at Q4_K_M fits your CPU only · DDR5 8-channel server entirely on the GPU at 8K context, at an estimated 19 tokens per second. There is room for its full 256K window.
Fully on GPU
8K context
Q4_K_M · 6.7 GB
Apache 2.0
Released 29 May 2026
Vision
The "unified" Gemma 4: text, image and audio in one 12B that fits a 12 GB card at Q4. 140+ languages.
The VRAM budget
weights 6.7 GB
Weights 6.7 GB
KV cache @ 8K 0.81 GB
Runtime overhead 0.6 GB
Free 196.6 GB of 204.8 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 22.4 GB | 23.8 GB | 256K | 5.8 | Reference | Long context |
| Q8_0 | 11.9 GB | 13.3 GB | 256K | 11 | −0.1% ppl | Long context |
| Q6_K | 9.2 GB | 10.6 GB | 256K | 14 | −0.4% ppl | Long context |
| Q5_K_M | 7.9 GB | 9.3 GB | 256K | 16 | −0.8% ppl | Long context |
| Q4_K_M | 6.7 GB | 8.2 GB | 256K | 19 | −1.9% ppl | Recommended |
| Q3_K_M | 5.5 GB | 6.9 GB | 256K | 24 | −5.4% ppl | Long context |
| Q2_K | 4.7 GB | 6.1 GB | 256K | 27 | −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. This model interleaves sliding-window layers (1024 tokens, 8 of 48 layers global), which is why its cache barely grows with context.
How to run it
$ llama-server \
-hf google/gemma-4-12B-it:Q4_K_M \
-c 8192 -ngl 99
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
01Download is 6.7 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02With no GPU, thread count matters more than clock. Start at one thread per physical core.
03There is room to go to the model's full 256K context on this card.