Yes — with 16.9 GB to spare
Mistral NeMo 12B at Q4_K_M fits your CPU only · DDR4 dual-channel entirely on the GPU at 8K context, at an estimated 3.1 tokens per second. Past 116K 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 · 6.9 GB
Apache 2.0
Released Jul 2024
Multilingual 12B with a 128K window, built with NVIDIA. A roleplay and fiction favourite that refuses to die.
The VRAM budget
weights 6.9 GB
Weights 6.9 GB
KV cache @ 8K 1.25 GB
Runtime overhead 0.6 GB
Free 16.9 GB of 25.6 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 22.7 GB | 24.6 GB | 14K | 0.9 | Reference | Fits |
| Q8_0 | 12.1 GB | 13.9 GB | 82K | 1.8 | −0.1% ppl | Long context |
| Q6_K | 9.3 GB | 11.2 GB | 100K | 2.3 | −0.4% ppl | Long context |
| Q5_K_M | 8.1 GB | 9.9 GB | 108K | 2.7 | −0.8% ppl | Long context |
| Q4_K_M | 6.9 GB | 8.7 GB | 116K | 3.1 | −1.9% ppl | Recommended |
| Q3_K_M | 5.6 GB | 7.4 GB | 124K | 3.8 | −5.4% ppl | Long context |
| Q2_K | 4.8 GB | 6.6 GB | 128K | 4.5 | −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 mistralai/Mistral-Nemo-Instruct-2407:Q4_K_M \
-c 8192 -ngl 99
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
01Download is 6.9 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 116K context on this card.