Model comparison
MN-GRAND-23.5B-Gutenberg-UNCENSORED-V2-GLM4.7-Thinking vs gemma-4-12B-it-qat-q4_0-unquantized
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-12B-it-qat-q4_0-unquantized's KV cache at 32K is 4.1× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| MN-GRAND-23.5B-Gutenberg-UNCENSORED-V2-GLM4.7-Thinking | gemma-4-12B-it-qat-q4_0-unquantized | |
|---|---|---|
| Parameters | 23.4B | 12.0B |
| Architecture | llama | gemma4 |
| Layers | 81 | 48 |
| Native context | 1,024,000 | 262,144 |
| Mixture of experts | no | no |
| Quantizations published | 33 | 2 |
| Smallest quantization | 5.00 GiB | 6.50 GiB |
| Q4_K_M | 13.37 GiB | — |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | MN-GRAND-23.5B-Gutenberg-UNCENSORED-V2-GLM4.7-Thinking | gemma-4-12B-it-qat-q4_0-unquantized | Ratio |
|---|---|---|---|
| 4,096 | 1.27 GiB | 0.72 GiB | 1.76× |
| 8,192 | 2.53 GiB | 0.97 GiB | 2.61× |
| 16,384 | 5.06 GiB | 1.47 GiB | 3.45× |
| 32,768 | 10.13 GiB | 2.47 GiB | 4.10× |
| 65,536 | 20.25 GiB | 4.47 GiB | 4.53× |
| 131,072 | 40.50 GiB | 8.47 GiB | 4.78× |