Yes — with 15.2 GB to spare
Qwen2.5-Coder 14B at Q4_K_M fits your CPU only · DDR4 dual-channel entirely on the GPU at 8K context, at an estimated 2.6 tokens per second. Past 88K 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 · 8.3 GB
Apache 2.0
Released Nov 2024
Noticeably better at whole-file edits than the 7B, still comfortable on 12 GB.
The VRAM budget
weights 8.3 GB
Weights 8.3 GB
KV cache @ 8K 1.50 GB
Runtime overhead 0.6 GB
Free 15.2 GB of 25.6 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 27.6 GB | 29.7 GB | — | ~0.8 | Reference | 4.1 GB over |
| Q8_0 | 14.6 GB | 16.7 GB | 55K | 1.5 | −0.1% ppl | Long context |
| Q6_K | 11.3 GB | 13.4 GB | 73K | 1.9 | −0.4% ppl | Long context |
| Q5_K_M | 9.8 GB | 11.9 GB | 81K | 2.2 | −0.8% ppl | Long context |
| Q4_K_M | 8.3 GB | 10.4 GB | 88K | 2.6 | −1.9% ppl | Recommended |
| Q3_K_M | 6.7 GB | 8.8 GB | 97K | 3.2 | −5.4% ppl | Long context |
| Q2_K | 5.8 GB | 7.9 GB | 102K | 3.7 | −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 Qwen/Qwen2.5-Coder-14B-Instruct:Q4_K_M \
-c 8192 -ngl 99
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
01Download is 8.3 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 88K context on this card.