Model comparison
TinyLlama-1.1B-Chat-v0.3 vs Qwen3-4B
At Q4_K_M, TinyLlama-1.1B-Chat-v0.3 is the smaller download — 667,822,976 bytes against 2,497,280,256. At long context the gap widens: TinyLlama-1.1B-Chat-v0.3's KV cache at 32K is 6.5× 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-v0.3 | Qwen3-4B | |
|---|---|---|
| Parameters | 1.1B | 4.0B |
| Architecture | llama | qwen3 |
| Layers | 22 | 36 |
| Native context | 2,048 | 40,960 |
| Mixture of experts | no | no |
| Quantizations published | 12 | 29 |
| Smallest quantization | 0.45 GiB | 1.01 GiB |
| Q4_K_M | 0.62 GiB | 2.33 GiB |
| Licence | apache-2.0 | — |
KV cache by context
the term that decides long-context viability
| Context | TinyLlama-1.1B-Chat-v0.3 | Qwen3-4B | Ratio |
|---|---|---|---|
| 4,096 | 0.09 GiB | 0.56 GiB | 6.55× |
| 8,192 | 0.17 GiB | 1.13 GiB | 6.55× |
| 16,384 | 0.34 GiB | 2.25 GiB | 6.55× |
| 32,768 | 0.69 GiB | 4.50 GiB | 6.55× |
| 65,536 | 1.38 GiB | 9.00 GiB | 6.55× |
| 131,072 | 2.75 GiB | 18.00 GiB | 6.55× |