Model comparison

OpenELM-3B-Instruct vs Qwen3-4B

At Q4_K_M, OpenELM-3B-Instruct is the smaller download — 1,887,788,096 bytes against 2,497,280,256.

From the file· summed bytes, KV per layer

Side by side

OpenELM-3B-InstructQwen3-4B
Parameters3.0B4.0B
Architectureopenelmqwen3
Layers36
Native context40,960
Mixture of expertsnono
Quantizations published1529
Smallest quantization1.07 GiB1.01 GiB
Q4_K_M1.76 GiB2.33 GiB
Licenceapple-amlr

KV cache by context

the term that decides long-context viability
ContextOpenELM-3B-InstructQwen3-4BRatio
4,0960.56 GiB
8,1921.13 GiB
16,3842.25 GiB
32,7684.50 GiB
65,5369.00 GiB
131,07218.00 GiB