README.md

---
license: other
license_name: modified-mit
license_link: LICENSE
base_model:
- moonshotai/Kimi-K2.6
---

# Model Overview

- **Model Architecture:** Kimi-K2.6
  - **Input:** Text, Image
  - **Output:** Text
- **Supported Hardware Microarchitecture:** AMD MI350/MI355
- **ROCm:** 7.2.0
- **PyTorch**: 2.9.1
- **Transformers**: 5.8.1
- **Operating System(s):** Linux
- **Inference Engine:** [SGLang](https://docs.sglang.ai/)/[vLLM](https://docs.vllm.ai/en/latest/)
- **Model Optimizer:** [AMD-Quark](https://quark.docs.amd.com/latest/index.html) (v0.11.1)
  - **Quantized layers:** `experts`, `shared_experts`
  - **Weight quantization:** OCP MXFP4, Static
  - **Activation quantization:** OCP MXFP4, Dynamic

This model was built with Kimi-K2.6 model by applying [AMD-Quark](https://quark.docs.amd.com/latest/index.html) for MXFP4 quantization.

# Model Quantization

The model was quantized from a BF16-decompressed version of [moonshotai/Kimi-K2.6](https://huggingface.co/moonshotai/Kimi-K2.6) using [AMD-Quark](https://quark.docs.amd.com/latest/index.html). The original checkpoint uses native INT4 (compressed-tensors) quantization; it was first decompressed to BF16 before applying MXFP4 quantization. The weights and activations are quantized to MXFP4.

**Quantization scripts:**
```
cd Quark/examples/torch/language_modeling/llm_ptq/
exclude_layers="*self_attn* *mlp.gate *lm_head *mm_projector* *vision_tower*"

python quantize_quark.py \
    --model_dir /path/to/Kimi-K2.6-bf16 \
    --quant_scheme mxfp4 \
    --exclude_layers $exclude_layers \
    --output_dir amd/Kimi-K2.6-MXFP4 \
    --model_export hf_format \
    --file2file_quantization
```

# Deployment
### Use with vLLM/SGLang

This model can be deployed efficiently using the [vLLM](https://docs.vllm.ai/en/latest/) and [SGLang](https://docs.sglang.ai/) backends.

## Evaluation
The model was evaluated on gsm8k benchmarks using the [vllm](https://github.com/vllm-project/vllm) framework.

### Accuracy

<table>
  <tr>
   <td><strong>Benchmark</strong>
   </td>
   <td><strong>Kimi-K2.6</strong>
   </td>
   <td><strong>Kimi-K2.6-MXFP4 (this model)</strong>
   </td>
   <td><strong>Recovery</strong>
   </td>
  </tr>
  <tr>
   <td>GSM8K (flexible-extract)
   </td>
   <td>0.9393
   </td>
   <td>0.9325
   </td>
   <td>99.3%
   </td>
  </tr>
</table>

### Reproduction

The GSM8K results were obtained using the vLLM framework, based on the Docker image `rocm/vllm-dev:nightly_main_20260417`, with lm-eval and amd-quark compiled and installed from source, and vLLM (version `0.19.1rc1.dev369+gb1dc87a09`) pre-installed in the docker image.

```
lm_eval \
  --model vllm \
  --model_args pretrained=amd/Kimi-K2.6-MXFP4,trust_remote_code=True,tensor_parallel_size=4 \
  --tasks gsm8k \
  --batch_size auto
```

# License
Modifications Copyright(c) 2026 Advanced Micro Devices, Inc. All rights reserved.