Model comparison
Marco-Nano-Instruct vs llama-3-youko-8b
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| Marco-Nano-Instruct | llama-3-youko-8b | |
|---|---|---|
| Parameters | 8.0B | 8.0B |
| Architecture | qwen3moe | llama |
| Layers | 28 | 32 |
| Native context | 32,768 | 8,192 |
| Mixture of experts | yes, 232 experts | no |
| Quantizations published | 36 | 2 |
| Smallest quantization | 2.44 GiB | 5.34 GiB |
| Q4_K_M | 5.00 GiB | — |
| Licence | apache-2.0 | llama3 |
KV cache by context
the term that decides long-context viability
| Context | Marco-Nano-Instruct | llama-3-youko-8b | Ratio |
|---|---|---|---|
| 4,096 | 0.44 GiB | 0.50 GiB | 1.14× |
| 8,192 | 0.88 GiB | 1.00 GiB | 1.14× |
| 16,384 | 1.75 GiB | 2.00 GiB | 1.14× |
| 32,768 | 3.50 GiB | 4.00 GiB | 1.14× |
| 65,536 | 7.00 GiB | 8.00 GiB | 1.14× |
| 131,072 | 14.00 GiB | 16.00 GiB | 1.14× |