Model Spend Arena2ND ED.
33B · dense · 2026-08-14

EXAONE 4.5 33B

Runs on ≤32 GB · Hugging Face ↗

llama.cppvLLM

Coding index 23.6 (Artificial Analysis)

LG AI Research's dense 33B, strong bilingual (Korean / English). Note its licence is research / non-commercial — read it before deploying.

Sizes on disk

Real GGUF file sizes = weight VRAM.

QuantizationSize
Q2_K12.4 GB
Q3_K_M16.1 GB
IQ4_XS18.0 GB
Q4_K_S19.0 GB
Q4_K_M20.0 GB
Q5_K_M23.5 GB
Q6_K27.1 GB
Q8_035.1 GB

Config tips

~21 GB at Q4, a 32 GB card. Good for Korean workloads; check the licence for any commercial use.