Is Atomight-V2.2-UltraThink-0.5B-OBLITERATED or Qwen3-0.6B smaller?

Qwen3-0.6B is smaller at Q4_K_M: 0.37 GiB against 0.37 GiB.

Which uses less memory at long context?

Atomight-V2.2-UltraThink-0.5B-OBLITERATED — its KV cache at 32K context is 9.3x smaller, because of how its attention layers cache.

Model comparison

Atomight-V2.2-UltraThink-0.5B-OBLITERATED vs Qwen3-0.6B

At Q4_K_M, Qwen3-0.6B is the smaller download — 396,705,472 bytes against 397,808,448. At long context the gap widens: Atomight-V2.2-UltraThink-0.5B-OBLITERATED's KV cache at 32K is 9.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Atomight-V2.2-UltraThink-0.5B-OBLITERATED	Qwen3-0.6B
Parameters	494M	752M
Architecture	qwen2	qwen3
Layers	24	28
Native context	32,768	40,960
Mixture of experts	no	no
Quantizations published	36	44
Smallest quantization	0.29 GiB	0.20 GiB
Q4_K_M	0.37 GiB	0.37 GiB
Licence	—	—

KV cache by context

the term that decides long-context viability

Context	Atomight-V2.2-UltraThink-0.5B-OBLITERATED	Qwen3-0.6B	Ratio
4,096	0.05 GiB	0.44 GiB	9.33×
8,192	0.09 GiB	0.88 GiB	9.33×
16,384	0.19 GiB	1.75 GiB	9.33×
32,768	0.38 GiB	3.50 GiB	9.33×
65,536	0.75 GiB	7.00 GiB	9.33×
131,072	1.50 GiB	14.00 GiB	9.33×