Model comparison
Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking vs gemma-4-12B-it
At Q4_K_M, Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking is the smaller download — 2,489,895,200 bytes against 7,121,861,440. At long context the gap widens: Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking's KV cache at 32K is 3.1× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking | gemma-4-12B-it | |
|---|---|---|
| Parameters | 4.3B | 12.0B |
| Architecture | gemma3 | gemma4 |
| Layers | 34 | 48 |
| Native context | 131,072 | 262,144 |
| Mixture of experts | no | no |
| Quantizations published | 12 | 59 |
| Smallest quantization | 1.61 GiB | 3.92 GiB |
| Q4_K_M | 2.32 GiB | 6.63 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking | gemma-4-12B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.25 GiB | 0.72 GiB | 2.90× |
| 8,192 | 0.33 GiB | 0.97 GiB | 2.97× |
| 16,384 | 0.48 GiB | 1.47 GiB | 3.04× |
| 32,768 | 0.79 GiB | 2.47 GiB | 3.11× |
| 65,536 | 1.42 GiB | 4.47 GiB | 3.15× |
| 131,072 | 2.67 GiB | 8.47 GiB | 3.17× |