README.md

---
base_model:
- moonshotai/Kimi-K2-Thinking
pipeline_tag: text-generation
---

[<img src="https://raw.githubusercontent.com/csabakecskemeti/devquasar/main/dq_logo_black-transparent.png" width="200"/>](https://devquasar.com)

'Make knowledge free for everyone'

Original INT4 model has been dequantized with my own custom script:

[DQ_int4-to-bf16_dequant](https://github.com/csabakecskemeti/DQ_int4-to-bf16_dequant)
(inspired by the deepseek V3 dequant script)

## Test/Proof

![kimi-think-proof](https://cdn-uploads.huggingface.co/production/uploads/64e6d37e02dee9bcb9d9fa18/EWnMFmGm9KcfePaFn9tgj.png)


### Zero Short Hexa-ball test, generated code by the Q3 quant produced:

![Kimi-Think_Hexa-Ball_test](https://cdn-uploads.huggingface.co/production/uploads/64e6d37e02dee9bcb9d9fa18/M24s1wupSCa-1u9nLcnBA.gif)

Quantized version of: [moonshotai/Kimi-K2-Thinking](https://huggingface.co/moonshotai/Kimi-K2-Thinking)
<a href='https://ko-fi.com/L4L416YX7C' target='_blank'><img height='36' style='border:0px;height:36px;' src='https://storage.ko-fi.com/cdn/kofi6.png?v=6' border='0' alt='Buy Me a Coffee at ko-fi.com' /></a>