Model comparison

llava-llama-3-8b-v1_1-transformers vs Muse-Glimmer-30B

At Q4_K_M, llava-llama-3-8b-v1_1-transformers is the smaller download — 4,921,098,752 bytes against 17,306,324,000.

From the file· summed bytes, KV per layer

Side by side

llava-llama-3-8b-v1_1-transformersMuse-Glimmer-30B
Parameters8.4B29.8B
Architecturellamamuse-glimmer
Layers52
Native context8,192131,072
Mixture of expertsnono
Quantizations published1536
Smallest quantization3.41 GiB8.31 GiB
Q4_K_M4.58 GiB16.12 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
Contextllava-llama-3-8b-v1_1-transformersMuse-Glimmer-30BRatio
4,0960.15 GiB
8,1920.20 GiB
16,3840.30 GiB
32,7680.50 GiB
65,5360.91 GiB
131,0721.72 GiB