Model comparison
Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-Thinking vs Qwen3-30B-A3B-Thinking-2507
At Q4_K_M, Qwen3-30B-A3B-Thinking-2507 is the smaller download — 18,556,685,824 bytes against 23,936,565,984.
From the file· summed bytes, KV per layer
Side by side
| Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-Thinking | Qwen3-30B-A3B-Thinking-2507 | |
|---|---|---|
| Parameters | 39.5B | 30.5B |
| Architecture | qwen35 | qwen3moe |
| Layers | 96 | 48 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 34 | 51 |
| Smallest quantization | 9.38 GiB | 7.05 GiB |
| Q4_K_M | 22.29 GiB | 17.28 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-Thinking | Qwen3-30B-A3B-Thinking-2507 | Ratio |
|---|---|---|---|
| 4,096 | 0.38 GiB | 0.38 GiB | 1.00× |
| 8,192 | 0.75 GiB | 0.75 GiB | 1.00× |
| 16,384 | 1.50 GiB | 1.50 GiB | 1.00× |
| 32,768 | 3.00 GiB | 3.00 GiB | 1.00× |
| 65,536 | 6.00 GiB | 6.00 GiB | 1.00× |
| 131,072 | 12.00 GiB | 12.00 GiB | 1.00× |