Is WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B or llama-3-youko-8b smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B — its KV cache at 32K context is 2.3x smaller, because of how its attention layers cache.

Model comparison

WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B vs llama-3-youko-8b

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B's KV cache at 32K is 2.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B	llama-3-youko-8b
Parameters	7.6B	8.0B
Architecture	qwen2	llama
Layers	28	32
Native context	32,768	8,192
Mixture of experts	no	no
Quantizations published	20	2
Smallest quantization	2.59 GiB	5.34 GiB
Q4_K_M	4.36 GiB	—
Licence	apache-2.0	llama3

KV cache by context

the term that decides long-context viability

Context	WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B	llama-3-youko-8b	Ratio
4,096	0.22 GiB	0.50 GiB	2.29×
8,192	0.44 GiB	1.00 GiB	2.29×
16,384	0.88 GiB	2.00 GiB	2.29×
32,768	1.75 GiB	4.00 GiB	2.29×
65,536	3.50 GiB	8.00 GiB	2.29×
131,072	7.00 GiB	16.00 GiB	2.29×