---
license: gemma
base_model:
- Vortex5/G4-Moonlight-Dusk-26B-A4B
---

# G4-Moonlight-Dusk-26B-A4B-APEX-GGUF

**APEX (Adaptive Precision for EXpert Models)** quantizations of [Vortex5/G4-Moonlight-Dusk-26B-A4B](https://huggingface.co/Vortex5/G4-Moonlight-Dusk-26B-A4B).

**I used the script from [LocalAI](https://github.com/mudler/LocalAI) team** | [APEX Project](https://github.com/mudler/apex-quant) | [Technical Report](https://github.com/mudler/apex-quant/blob/main/paper/APEX_Technical_Report.pdf)

**I also used the latest chat_template.jinja(2026-07-22 , if google ever update the template again , override the template with the latest) from [Google](https://huggingface.co/google/gemma-4-26B-A4B-it/blob/main/chat_template.jinja) , which fixes a lot of issue , for RP specifically thinking will not longer leak out .**

**imatrix data is from [mradermacher](https://huggingface.co/mradermacher/G4-Moonlight-Dusk-26B-A4B-i1-GGUF/blob/main/G4-Moonlight-Dusk-26B-A4B.imatrix.gguf)**