# Credits

## Qwen Team / Alibaba
`Qwen/Qwen3.8-27B`

Foundation architecture/model generation.

## Blackfrost-AI
`Blackfrost-AI/Qwen3.8-27B-ABLITERATED-BF16`

Abliterated BF16 Qwen3.8 checkpoint used as the actual merge foundation for this custom release.

## TeichAI
`TeichAI/Qwen3.8-27B-Fable-Distill-LoRA`

Native Qwen3.8 Fable5-Distill adapter used in this build.

TeichAI's adapter model tree explicitly lists `Qwen/Qwen3.8-27B` as its base.

## Fable datasets credited by TeichAI

- `armand0e/claude-fable-5-claude-code`
- `armand0e/Fable-5-Chat`

TeichAI also reports using a larger private corpus of personal Fable 5 data.

## Benchmarks
TeichAI credits **@nightmedia** for the published benchmark measurements.

## Training stack
TeichAI credits:

- Unsloth
- Hugging Face TRL

## GGUF ecosystem
ggml-org / llama.cpp

Used for GGUF conversion, importance-matrix generation, quantization and inference.

## This derivative

The publisher of this repository - verkiki - is responsible for:

- applying the TeichAI native Qwen3.8 LoRA to the Blackfrost Qwen3.8 abliterated BF16 weights;
- conversion to GGUF;
- creation of the final post-merge importance matrix;
- IQ4_XS quantization;
- local validation and publication.

This repository is an independent derivative and is not endorsed by the upstream projects.