Question 1

Can the NVIDIA L4 run GLM-5.3-Flash (Ox Alpha)?

Accepted Answer

GLM-5.3-Flash (Ox Alpha) needs about 166GB working set (weights 164GB + KV cache 2GB at 4k context). The NVIDIA L4 offers ~24GB usable — verdict: no. 166.0 GB vs 24 GB usable — does not fit

Question 2

How fast is GLM-5.3-Flash (Ox Alpha) on the NVIDIA L4?

Accepted Answer

~18–29 tok/s (active-expert estimate) on the NVIDIA L4. computed roofline: 300 GB/s × 0.55–0.88 efficiency window / 9.2GB (Q4_K_M) — bands, never points. MoE: file is 164GB (that decides fit) but only ~18B of experts activate per token (~9.2GB streamed) — band below is the active-expert estimate, optimistic; community MoE rows run several× the naive bandwidth math

Question 3

What should I run GLM-5.3-Flash (Ox Alpha) on instead?

Accepted Answer

ASUS ExpertCenter Pro ET900N G3 (GB300) runs it fully ($60000).

Can the NVIDIA L4 run GLM-5.3-Flash (Ox Alpha)?

Run GLM-5.3-Flash (Ox Alpha) on instead