Verkiki/Qwen3.8-27b-Fable5-Distill-Abliterated-GGUF

Qwen3.8-27B Abliterated × Fable5-Distill — IQ4_XS - MTP ON

Abliterated Qwen3.8-27B + native Qwen3.8 Fable5-Distill behavior + post-merge imatrix IQ4_XS: a compact local build aimed at planning, coding, reasoning and agentic execution.

This release combines:

• —Foundation weights: Blackfrost-AI/Qwen3.8-27B-ABLITERATED-BF16
• —LoRA: TeichAI/Qwen3.8-27B-Fable-Distill-LoRA
• —LoRA intended base: Qwen/Qwen3.8-27B
• —Architecture: qwen3_5, 27B class
• —Final format: GGUF
• —Target quant: IQ4_XS
• —Calibration: post-merge imatrix

This is not a cross-version Qwen3.6 → Qwen3.8 transplant. The TeichAI adapter is a native Qwen3.8-27B LoRA.

The only deliberate base variation is that the adapter is merged onto the Blackfrost abliterated derivative of Qwen3.8, rather than the untouched official Qwen3.8 checkpoint.

Why this build exists

The goal is straightforward: take the behavioral gains of TeichAI's Fable-distilled Qwen3.8 tune, preserve the lower-refusal character of the Blackfrost Qwen3.8 derivative, and package the result in a compact local format.

The result is aimed at users who want a 27B local model with strong planning, coding, long-form reasoning and agentic behavior, but without the memory footprint of BF16 or Q8.

Fable5-Distill: measured gains on Qwen3.8

TeichAI published direct benchmark results for its Qwen3.8-27B-Fable5-Distill model:

These are unusually clean results for a behavioral distill: the Fable tune improves all three published evaluation scores rather than trading one off against another.

TeichAI describes this as a light tune on Qwen3.8, trained using the cited public Fable 5 chat/agent datasets plus a substantially larger private corpus of personal Fable 5 data.

The adapter's Hugging Face model tree explicitly lists:

Base model: Qwen/Qwen3.8-27B
→ TeichAI/Qwen3.8-27B-Fable-Distill-LoRA

So this release keeps the LoRA on the same Qwen3.8 model generation for which it was trained.

Source: TeichAI/Qwen3.8-27B-Fable-Distill-LoRA TeichAI/Qwen3.8-27B-Fable-Distill

What Fable5-Distill adds

Based on TeichAI's model card and its direct Qwen3.8 benchmarks, the expected behavioral shift is toward:

• —stronger task decomposition;
• —more deliberate planning;
• —more agent-like execution;
• —better instruction retention across multi-step work;
• —stronger general reasoning signal on ARC;
• —improved straightforward factual/semantic decision accuracy on BoolQ;
• —preservation of Qwen3.8's thinking controls.

TeichAI's Qwen3.8 release explicitly supports:

enable_thinking
reasoning_effort = low | medium | xhigh

with the template defaulting to xhigh.

That makes this build attractive for coding agents, technical problem-solving, autonomous task planning, research workflows, and long-form implementation work.

Why the Blackfrost base

This model does not merge the Fable LoRA into vanilla Qwen3.8.

Instead, it starts from:

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

and merges the native Qwen3.8 Fable LoRA into those BF16 weights.

The intended effect is to combine:

Blackfrost Qwen3.8
    ↓
reduced refusal tendency

+

TeichAI Qwen3.8 Fable5-Distill LoRA
    ↓
measured reasoning / instruction benchmark gains
    ↓
Fable-style planning and execution behavior

The resulting merged checkpoint is a new derivative and should be evaluated on its own. A benchmark from either parent is evidence about lineage, not automatically a measured score for the final IQ4_XS.

Expected strengths

Agentic coding

Qwen3.8 is designed around modern agentic workloads, and the Fable distill is trained on Fable 5 chat and agent traces.

The combination is particularly well suited to:

• —planning code changes before editing;
• —multi-file repository work;
• —debugging;
• —iterative implementation;
• —tool-oriented workflows;
• —implementation plans with dependencies and checkpoints.

Planning

This is the headline use case.

Fable-style data rewards a model for maintaining goals over multiple steps instead of immediately producing the first plausible answer.

Expect the model to be most interesting on prompts such as:

• —“inspect → plan → implement → verify”;
• —architecture decisions;
• —refactoring;
• —research plans;
• —multi-step technical tasks;
• —agent loops.

Reasoning

The TeichAI Qwen3.8 Fable model improved ARC Challenge from 0.591 to 0.637 and ARC Easy from 0.782 to 0.832.

That is concrete evidence that the tune does more than merely imitate a writing style.

Instruction comprehension

BoolQ improves from 0.896 to 0.911 in TeichAI's published comparison.

That supports the expectation that the Fable tune retains — and in this evaluation improves — basic instruction/question understanding.

Low-friction local use

The Blackfrost foundation was selected specifically for its abliterated / reduced-refusal character.

This can make the final model more useful for legitimate technical and research workflows that trigger unnecessary false-positive refusals in more conservative checkpoints.

IQ4_XS + post-merge imatrix

The final build uses IQ4_XS with a custom importance matrix generated after the Blackfrost Abliterated + TeichAI Fable5-Distill merge.

This is the exact imatrix artifact included in this repository:

filename:            Qwen3.8-27B-Abliterated-Fable5-Distill-imatrix.gguf
GGUF version:        3
general.type:        imatrix
GGUF tensor count:   992
imatrix entries:     496
chunk_count:         280
chunk_size:          512
token positions:     143,360
dataset recorded:    C:\AI\calibration.txt
file size:           13,642,624 bytes
SHA-256:             82ca0c264e505aae0dc391f31d8e447281085cee8ab4314c835ae4edc18b5db2

The quantizer log for this build confirmed:

llama_model_quantize_impl: have importance matrix data with 496 entries

Calibration volume recorded in the file:

280 × 512 = 143,360 token positions

This is calibration metadata, not a benchmark score. The imatrix guides llama.cpp's importance-aware quantization so that more sensitive activation directions receive preferential treatment during IQ4_XS compression.

The raw calibration corpus is not included here. The imatrix records its source as:

C:\AI\calibration.txt

Not a blind 4-bit conversion

Blackfrost Qwen3.8 Abliterated BF16
        +
TeichAI Qwen3.8 Fable5-Distill LoRA
        ↓
merged high-precision Qwen3.8
        ↓
post-merge calibration
        ↓
496-entry importance matrix
        ↓
IQ4_XS

llama.cpp may intentionally keep selected sensitive tensors at higher precision. In the actual quantization log for this build, output.weight was converted to Q6_K while the overall model remained an IQ4_XS-class quant. This mixed-precision behavior is expected.

Multimodal / vision

TeichAI's Qwen3.8 Fable5-Distill is multimodal.

Their published GGUF build uses a separate mmproj and notes:

• —F16 projector: usual/default choice;
• —BF16 projector: matches source dtype;
• —F32 projector: fully unquantized projector.

This repository's primary IQ4_XS GGUF is the language model component.

For vision, publish and use a matching Qwen3.8 mmproj that you have actually tested.

Example:

llama-server \
  -m Qwen3.8-27B-Abliterated-Fable5-Distill-IQ4_XS.gguf \
  --mmproj mmproj-F16.gguf

MTP / NextN

The official TeichAI Qwen3.8 Fable GGUF release retains the Multi-Token Prediction head:

64 transformer layers
+ 1 MTP / nextn layer
= qwen35.block_count 65

TeichAI keeps that MTP head in BF16 in its own quantizations.

If this custom release is converted with MTP removed (--no-mtp / --no-nextn), do not claim MTP speculative decoding support.

The final repository should state the exact build choice after the final GGUF is verified.

Sampling

TeichAI lists the Qwen3.8 base recommendations as:

temperature = 1.0
top_p       = 0.95
top_k       = 20

Thinking controls:

enable_thinking = true/false
reasoning_effort = low | medium | xhigh

For coding and agent workflows, medium is a practical place to start; use xhigh when the task benefits from longer deliberation.

llama.cpp

llama-cli \
  -m Qwen3.8-27B-Abliterated-Fable5-Distill-IQ4_XS.gguf \
  -c 8192

Server:

llama-server \
  -m Qwen3.8-27B-Abliterated-Fable5-Distill-IQ4_XS.gguf \
  -c 8192

LM Studio

Import:

lms import "Qwen3.8-27B-Abliterated-Fable5-Distill-IQ4_XS.gguf"

Then configure GPU offload according to available VRAM.

Benchmark integrity

This repository makes a strict distinction between measured parent-model evidence and measured final-model evidence.

Directly measured by TeichAI on Qwen3.8 Fable5-Distill

• —ARC Challenge: 0.637
• —ARC Easy: 0.832
• —BoolQ: 0.911

Baseline Qwen3.8 numbers from the same TeichAI comparison

• —ARC Challenge: 0.591
• —ARC Easy: 0.782
• —BoolQ: 0.896

Not yet claimed for this custom Blackfrost + Fable IQ4_XS

• —those exact benchmark scores;
• —final imatrix statistics;
• —final refusal rate;
• —final vision benchmark;
• —final long-context benchmark.

Once the published IQ4_XS is benchmarked directly, its results should be added in a separate Final Build Benchmarks section.

Credits

• —Qwen Team / Alibaba — Qwen/Qwen3.8-27B
• —Blackfrost-AI — Blackfrost-AI/Qwen3.8-27B-ABLITERATED-BF16
• —TeichAI — TeichAI/Qwen3.8-27B-Fable-Distill-LoRA
• —armand0e — claude-fable-5-claude-code and Fable-5-Chat
• —@nightmedia — benchmark work credited by TeichAI
• —Unsloth + Hugging Face TRL — training stack credited by TeichAI
• —ggml-org / llama.cpp — GGUF, imatrix, IQ4_XS and runtime tooling

This is an independent derivative and is not an official release or endorsement by any of the above projects.

Sources

See SOURCES.md.