Model comparison

starcoder2-3b vs Llama-3.1-8B-Instruct

At Q4_K_M, starcoder2-3b is the smaller download — 1,848,976,448 bytes against 4,920,739,168.

From the file· summed bytes, KV per layer

Side by side

starcoder2-3bLlama-3.1-8B-Instruct
Parameters3.0B8.0B
Architecturestarcoder2llama
Layers3032
Native context16,384131,072
Mixture of expertsnono
Quantizations published3745
Smallest quantization1.07 GiB2.02 GiB
Q4_K_M1.72 GiB4.58 GiB
Licencebigcode-openrail-mllama3.1

KV cache by context

the term that decides long-context viability
Contextstarcoder2-3bLlama-3.1-8B-InstructRatio
4,0960.50 GiB
8,1921.00 GiB
16,3842.00 GiB
32,7684.00 GiB
65,5368.00 GiB
131,07216.00 GiB