Model comparison

TinyLlama-1.1B-Chat-v1.0 vs Llama-3.2-1B-Instruct

At Q4_K_M, TinyLlama-1.1B-Chat-v1.0 is the smaller download — 667,814,880 bytes against 807,690,688. At long context the gap widens: TinyLlama-1.1B-Chat-v1.0's KV cache at 32K is 1.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

TinyLlama-1.1B-Chat-v1.0Llama-3.2-1B-Instruct
Parameters1.1B1.2B
Architecturellamallama
Layers2216
Native context2,048131,072
Mixture of expertsnono
Quantizations published2639
Smallest quantization0.45 GiB0.39 GiB
Q4_K_M0.62 GiB0.75 GiB
Licenceapache-2.0—

KV cache by context

the term that decides long-context viability
ContextTinyLlama-1.1B-Chat-v1.0Llama-3.2-1B-InstructRatio
4,0960.09 GiB0.13 GiB1.45×
8,1920.17 GiB0.25 GiB1.45×
16,3840.34 GiB0.50 GiB1.45×
32,7680.69 GiB1.00 GiB1.45×
65,5361.38 GiB2.00 GiB1.45×
131,0722.75 GiB4.00 GiB1.45×