No — not on this device
GLM-5.3 744B-A40B at Q4_K_M needs 424.7 GB against 8.8 GB usable, and the shortfall of 415.9 GB is more than 16 GB of system RAM can cover at a tolerable speed. A smaller sibling or a lower quantisation is the honest answer here.
Same base as GLM-5.2, new post-training: Z.ai's most capable open-weights coder (Terminal-Bench 3.0 28.3, DeepSWE 66.9). Not MIT any more — its own GLM-5.3 licence. Same ceiling as 5.2: 512 GB of unified memory at Q4.
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 745.1 GB | 746.4 GB | — | ~0.4 | −0.1% ppl | 737.6 GB over |
| Q6_K | 575.1 GB | 576.3 GB | — | ~0.5 | −0.4% ppl | 567.5 GB over |
| Q5_K_M | 497.0 GB | 498.3 GB | — | ~0.5 | −0.8% ppl | 489.5 GB over |
| Q4_K_M | 423.4 GB | 424.7 GB | — | ~0.6 | −1.9% ppl | 415.9 GB over |
| Q3_K_M | 342.8 GB | 344.0 GB | — | ~0.8 | −5.4% ppl | 335.2 GB over |
| Q2_K | 293.7 GB | 294.9 GB | — | ~0.9 | −15% ppl | 286.1 GB over |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model uses multi-head latent attention, so its cache is a compressed latent rather than full K and V.
How to run it
$ llama-server \
-hf zai-org/GLM-5.3:Q4_K_M \
-c 8192 -ngl 1
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.