tags:
- qwen3.5
- text-generation
- quanto
- int8
library_name: transformers
license: other
Qwen3.5 9B Abliterated v2 Quanto BF16/INT8 Mirror
This repository provides the model files used by the local liuliu Faithful H3 Web tool.
Provenance
The checkpoint and companion tokenizer/configuration files are distributed for local use with the linked tool. This repository does not claim authorship or ownership of the model.
- Intended application: https://github.com/wodeshijie1234/faithful-h3-web
Checkpoint identity
- File:
Qwen3.5-9B-Abliterated_v2_quanto_bf16_int8.safetensors - Size:
8,957,488,932bytes - SHA256:
eb03df5ccba4536eb64cf096c08b068eb84cfd2d2aa798cd45f31a0f67e339e6
The application verifies both the exact file size and SHA256 before loading the checkpoint. The older non-v2 checkpoint is not interchangeable and is not included here.
Runtime
The Faithful H3 application loads the text-only Qwen3.5 9B checkpoint with Transformers and Optimum Quanto on a CUDA-capable NVIDIA GPU. The repository includes the tokenizer, chat template, vocabulary, and model configuration needed by that runtime.
GGUF variant
An optional llama.cpp GGUF variant is also available for the lightweight local backend.
- File:
Qwen3.5-9B-Abliterated-text-Q4_K_M_bis.gguf - Size:
5,627,044,704bytes - SHA256:
dba64d0e5cce0739e27535ee0a6b75249eb8006ce8b2d6c060e20750035c4695
The GGUF file is a separately quantized distribution. It does not replace the existing Quanto checkpoint or its companion files.
License and responsible use
This is a third-party model artifact. Review the Qwen model terms, base-model licenses, and all applicable laws before using, modifying, or redistributing it. The license: other metadata is deliberately conservative because this repository does not grant additional model rights.