Model comparison

Qwen2.5-3B-Instruct-abliterated vs Qwen3-8B

At Q4_K_M, Qwen2.5-3B-Instruct-abliterated is the smaller download — 4,034,837,120 bytes against 5,027,783,488. At long context the gap widens: Qwen2.5-3B-Instruct-abliterated's KV cache at 32K is 4.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen2.5-3B-Instruct-abliteratedQwen3-8B
Parameters3.1B8.2B
Architectureqwen2qwen3
Layers3636
Native context32,76840,960
Mixture of expertsnono
Quantizations published4051
Smallest quantization1.21 GiB2.12 GiB
Q4_K_M3.76 GiB4.68 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen2.5-3B-Instruct-abliteratedQwen3-8BRatio
4,0960.14 GiB0.56 GiB4.00×
8,1920.28 GiB1.13 GiB4.00×
16,3840.56 GiB2.25 GiB4.00×
32,7681.13 GiB4.50 GiB4.00×
65,5362.25 GiB9.00 GiB4.00×
131,0724.50 GiB18.00 GiB4.00×