Model comparison

Llama-3.1-Nemotron-70B-Instruct-HF-abliterated vs Qwen3.8-27B

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 42,520,398,816.

From the file· summed bytes, KV per layer

Side by side

Llama-3.1-Nemotron-70B-Instruct-HF-abliteratedQwen3.8-27B
Parameters70.6B27.8B
Architecturellamaqwen35
Layers64
Native context262,144
Mixture of expertsnono
Quantizations published3722
Smallest quantization14.29 GiB8.39 GiB
Q4_K_M39.60 GiB15.66 GiB
Licencellama3.1apache-2.0

KV cache by context

the term that decides long-context viability
ContextLlama-3.1-Nemotron-70B-Instruct-HF-abliteratedQwen3.8-27BRatio
4,0960.25 GiB
8,1920.50 GiB
16,3841.00 GiB
32,7682.00 GiB
65,5364.00 GiB
131,0728.00 GiB