Model comparison
gpt2 vs SmolLM2-135M-Instruct
At Q4_K_M, SmolLM2-135M-Instruct is the smaller download — 105,454,144 bytes against 112,858,624.
From the file· summed bytes, KV per layer
Side by side
| gpt2 | SmolLM2-135M-Instruct | |
|---|---|---|
| Parameters | 137M | 135M |
| Architecture | gpt2 | llama |
| Layers | 12 | 30 |
| Native context | — | 8,192 |
| Mixture of experts | no | no |
| Quantizations published | 49 | 41 |
| Smallest quantization | 0.06 GiB | 0.08 GiB |
| Q4_K_M | 0.11 GiB | 0.10 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gpt2 | SmolLM2-135M-Instruct | Ratio |
|---|---|---|---|
| 4,096 | — | 0.09 GiB | — |
| 8,192 | — | 0.18 GiB | — |
| 16,384 | — | 0.35 GiB | — |
| 32,768 | — | 0.70 GiB | — |
| 65,536 | — | 1.41 GiB | — |
| 131,072 | — | 2.81 GiB | — |