---
base_model:
- Qwen/Qwen3.5-122B-A10B
---
## Updates
- 5/18/2026: I've uploaded new quants that include the MTP Tensors (@ Q8_0).
- 3/10/2026: I've uploaded new quants using the new fused Up + Gate conversion, this offers up to a +10% boost in prompt processing speed from my testing.

## Description
This repo contains specialized MoE-quants for Qwen3.5-122B-A10B. The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization. To that end, the quantization type default is kept in high quality and the FFN UP + FFN GATE tensors are quanted down along with the FFN DOWN tensors.

| Quant   | Size                  | Mixture                            | PPL                 | 1-(Mean PPL(Q)/PPL(base)) | KLD                 |
| :------ | :-------------------- | :--------------------------------- | :------------------ | :------------------------ | :------------------ |
| Q8_0    | 123.44 GiB (8.51 BPW) | Q8_0                               | 4.817202 ± 0.028383 | +0.0132%                  | 0.003675 ± 0.000034 |
| Q5_K_M  | 87.74 GiB (6.05 BPW)  | Q8_0 / Q5_K / Q5_K / Q6_K          | 4.823697 ± 0.028442 | +0.1481%                  | 0.005549 ± 0.000042 |
| Q4_K_M  | 73.96 GiB (5.10 BPW)  | Q8_0 / Q4_K / Q4_K / Q5_K          | 4.829970 ± 0.028461 | +0.2783%                  | 0.010420 ± 0.000078 |
| IQ4_XS  | 58.77 GiB (4.05 BPW)  | Q8_0 / IQ3_S / IQ3_S / IQ4_XS      | 4.914083 ± 0.028949 | +2.0246%                  | 0.027803 ± 0.000206 |
| IQ3_S   | 45.87 GiB (3.16 BPW)  | Q6_K / IQ2_S / IQ2_S / IQ3_S       | 5.128301 ± 0.030530 | +6.4722%                  | 0.074536 ± 0.000521 |
| IQ2_XXS | 34.08 GiB (2.35 BPW)  | Q4_K / IQ2_XXS / IQ2_XXS / IQ2_XXS | 5.733499 ± 0.035073 | +19.0371%                 | 0.185455 ± 0.001112 |

![kld_graph](kld_data/01_kld_vs_filesize.png "Chart showing Pareto KLD analysis of quants")
![ppl_graph](kld_data/02_ppl_vs_filesize.png "Chart showing Pareto PPL analysis of quants")