---

base_model:

- nightmedia/Qwen3.8-27B-Fable-Fusion-F711-GAIN-mxfp4-mlx

base_model_relation: quantized

pipeline_tag: image-text-to-text

license: other

tags:

- qwen3.8

- qwen3.5

- gguf

- llama.cpp

- iq4-xs

- imatrix

- mtp

- speculative-decoding

- multimodal

- vision

- merge

- 27b

---

FableFusion GAIN F711 27B

Qwen3.8-class fusion · IQ4_XS · model-specific imatrix · embedded MTP · multimodal

> Not a generic Qwen quant. This release starts from Nightmedia's Fable-Fusion F711 GAIN MXFP4 checkpoint — a high-performing experimental fusion whose published MXFP4 source measurements reached 0.690 ARC, 3.803 ± 0.023 perplexity, and 185 tok/s. The exact MXFP4 checkpoint was then structurally recovered, dequantized into an ordinary FP16 bridge representation, calibrated with a model-specific importance matrix, converted to IQ4_XS, and completed with an embedded Qwen3.8-compatible MTP/NextN head.

The result is a compact GGUF release built around a model lineage that already showed a large measurable jump over stock Qwen3.8 in Nightmedia's own same-family MXFP4 tests — while preserving Qwen3.8's 27B / 64-layer multimodal architecture and local-runtime flexibility.

Link for script to recover mxfp4 to gguf: https://github.com/verkiki/quantbridge - this version was not tested yet; an older version was used to generate the f16.

Link for script to extract mtp from donor and put into the new gguf: https://github.com/verkiki/mtp-forge - this version was not tested yet; an older version was used to generate the final IQ4_XS.

---

Why this model is worth downloading

A source lineage with measurable gains

Nightmedia's published MXFP4 measurements show a clear progression from stock Qwen3.8 through GAIN, Fable, and finally F711:

| Published MXFP4 model | ARC | Perplexity | Peak memory | tok/s |

|---|---:|---:|---:|---:|

| Qwen3.8-27B baseline | 0.581 | 5.952 ± 0.051 | 21.30 GB | 148 |

| Cold-Fusion GAIN V1.1 | 0.644 | 4.042 ± 0.026 | 21.30 GB | 185 |

| Cold-Fusion GAIN V1.1 + Fable | 0.654 | 3.975 ± 0.025 | 21.30 GB | 186 |

| Fable-Fusion F711 GAIN | 0.690 | 3.803 ± 0.023 | 21.30 GB | 185 |

Against the published stock Qwen3.8 MXFP4 baseline, the F711 source shows:

• +0.109 ARC absolute — +18.8% relative
• 36.1% lower perplexity
• 25.0% higher reported throughput
• the same reported 21.30 GB peak-memory figure in Nightmedia's MXFP4 test setup

That is the performance signal this release is built on.

> Benchmark honesty: those are measurements of Nightmedia's upstream MXFP4 F711 source, not newly measured scores for this IQ4_XS binary. This release does not relabel inherited results as fresh benchmarks.

Fable contributes measurable signal too

The Fable component is not just a name in the merge tree. TeichAI reports the Qwen3.8 Fable Distill moving from the Qwen3.8 baseline to:

| Model | ARC Challenge | ARC Easy | BoolQ |

|---|---:|---:|---:|

| Qwen3.8-27B | 0.591 | 0.782 | 0.896 |

| Qwen3.8-27B-Fable-Distill | 0.637 | 0.832 | 0.911 |

That corresponds to approximately +7.8% relative ARC, +6.4% relative ARC Easy, and +1.7% relative BoolQ in the published Fable evaluation.

Qwen3.8 foundation, not a stripped text-only derivative

Qwen3.8-27B is a dense 27B, 64-layer native vision-language architecture with a 262,144-token native context window, extensible by Qwen to 1,000,000 tokens, and a foundation trained with Multi-Token Prediction. It supports image/video understanding and flexible reasoning control at the architecture/template level.

This release keeps the multimodal model path. Vision is used through the matching mmproj GGUF.

---

What was actually done to build this release

This is important: the available Nightmedia source was already MXFP4.

This release was not produced from an unreleased original BF16 checkpoint. The FP16 checkpoint used later in the workflow was a reconstruction of the exact MXFP4 source values and served as a bridge format for GGUF tooling.

Nightmedia Fable-Fusion F711 GAIN
BF16 merge lineage
        ↓
Nightmedia MLX MXFP4 release
        ↓
structural shard validation
        ↓
safe tensor I/O + MLX MXFP4 dequantization
        ↓
ordinary FP16 reconstruction of the available MXFP4 values
        ↓
model-specific importance-matrix calibration
        ↓
IQ4_XS GGUF backbone
        ↓
embedded Qwen3.8-compatible MTP/NextN graft
        ↓
FableFusion GAIN F711 27B — final GGUF

Recovery audit — real numbers from the conversion

The local source checkpoint contained 3 Safetensors shards:

• shard 1: 5,348,727,334 bytes
• shard 2: 5,349,386,411 bytes
• shard 3: 4,515,944,339 bytes
• total source shard payload: 15,214,058,084 bytes / 14.169 GiB

All three shard headers and tensor byte ranges were validated before conversion.

The source tensor plan contained:

• 1,682 source tensor entries
• 498 packed MXFP4 weights
• 498 MXFP4 scale tensors
• 1,184 reconstructed ordinary model tensors
• 333 vision-related tensors preserved
• 0 physical MTP tensors in the Nightmedia MXFP4 source

The source quantization metadata was:

• mode: MXFP4
• bits: 4
• group size: 32
• packed weight dtype: uint32
• scale dtype: uint8

Before the bulk conversion, a real attention projection was tested end-to-end:

packed weight: (1024, 640) uint32
scales:        (1024, 160) uint8
                    ↓ MXFP4 dequantization
FP16 weight:   (1024, 5120)

This is why the release is described as an MXFP4-recovered IQ4_XS, not as a simple "BF16 → Q4" conversion.

Windows-safe recovery path

The source files were not corrupt. Standard Safetensors access successfully validated and exposed all source entries, while MLX's direct large-file loading path failed on Windows. The recovery therefore used normal Safetensors I/O for the files and MLX specifically for the MXFP4 dequantization math.

Plain BF16 tensors that could not be exposed through NumPy were handled from their serialized payload and converted to FP16 explicitly.

Model-specific imatrix

The importance matrix was generated from a calibration representation of this recovered F711 backbone rather than copied from an unrelated public quant.

The final IQ4_XS therefore uses calibration data tied to the same recovered model being released.

Embedded MTP after backbone quantization

The Nightmedia MXFP4 source declared MTP in configuration but contained no physical MTP tensor block in the audited weights.

The final GGUF therefore adds an embedded compatible Qwen3.8 NextN/MTP block after the F711 backbone quantization. The donor block was copied in GGUF form rather than being treated as part of the recovered Nightmedia source.

The result is a single main GGUF designed for current llama.cpp builds that support:

--spec-type draft-mtp

No separate draft-model file is required when the embedded MTP block is recognized by the runtime.

---

The real model lineage

The final Nightmedia F711 source is a merge of merges, not a two-model fine-tune.

Qwen3.8 GAIN + Fable branch

Nightmedia documents Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-Fable as a BF16 NuSLERP merge of:

DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1   weight 1.6
armand0e/Qwen3.8-27B-Fable-Distill          weight 0.4
merge method: NuSLERP
dtype: bfloat16

The Fable model is currently published under the TeichAI account as TeichAI/Qwen3.8-27B-Fable-Distill; Nightmedia's recipe records the component under the armand0e name.

DavidAU describes the Cold Fusion model as a light, focused Qwen3.8 tune built with the GAIN + Unsloth training approach, targeting stronger general capability and shorter reasoning traces.

Qwen3.6 Architect / Polaris / Fable / F451 branch

Nightmedia's Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 is itself a BF16 NuSLERP merge:

Qwen3.6-27B-Architect-Polaris2-Fable-B       weight 1.4
Qwen3.6-27B-Architect-Polaris-Fable-F451     weight 0.6
merge method: NuSLERP
dtype: bfloat16

Nightmedia reports for this branch, at MXFP4:

ARC      0.701
ARC Easy 0.873
BoolQ    0.909
HellaSwag 0.786
OpenBookQA 0.488
PIQA     0.813
WinoGrande 0.759
PPL      3.854 ± 0.024
tok/s    197

Final F711 fusion

Nightmedia then combines the Qwen3.8 GAIN+Fable branch with the Qwen3.6 Architect/Polaris/Fable/F451 branch:

Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-Fable     weight 1.4
Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 weight 0.6
merge method: NuSLERP
dtype: bfloat16

That merge became the Fable-Fusion F711 GAIN lineage that Nightmedia later released in MLX MXFP4 form — the exact source recovered for this GGUF.

Condensed tree

Qwen/Qwen3.8-27B
│
├── DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1
│
└── armand0e / TeichAI Qwen3.8-27B-Fable-Distill
          │
          └── NuSLERP BF16 (1.6 / 0.4)
                    ↓
      Nightmedia Qwen3.8 Cold-Fusion-GAIN-V1.1-Fable
                    │
                    │  weight 1.4
                    │
                    ├───────────────────────────────────────┐
                                                            │
Qwen3.6 Architect / Polaris / Fable / F451 lineage          │
│                                                           │
├── Architect-Polaris2-Fable-B                              │
└── Architect-Polaris-Fable-F451                            │
          │                                                 │
          └── NuSLERP BF16 (1.4 / 0.6)                      │
                    ↓                                       │
      Nightmedia Qwen3.6 Architect-Polaris2-Fable-B-F451    │
                    │                                       │
                    │ weight 0.6                            │
                    └───────────────────────────────────────┘
                                  ↓
                         NuSLERP BF16
                                  ↓
                Nightmedia Fable-Fusion F711 GAIN
                                  ↓
                           MLX MXFP4 release
                                  ↓
            Verkiki: recovery + custom imatrix + IQ4_XS
                                  ↓
                    embedded MTP / NextN block
                                  ↓
                    FableFusion GAIN F711 27B

---

What to expect

This release is aimed at people who want one strong 27B local model that does more than generic chat:

• reasoning and long-form problem solving inherited from Qwen3.8 and the GAIN branch;
• creative writing / narrative behavior inherited from the Fable line;
• coding, research and agentic capability from the Qwen3.8 foundation and the Architect/Polaris branch;
• native multimodal architecture when paired with the matching mmproj;
• flexible thinking behavior when supported by the runtime/chat template;
• embedded MTP speculative decoding support in compatible llama.cpp builds;
• a compact IQ4_XS backbone calibrated with a model-specific importance matrix.

The strongest quantitative evidence available before a dedicated benchmark of this exact IQ4_XS binary is the upstream F711 source benchmark ladder shown above. The final IQ4_XS should be evaluated separately if you need apples-to-apples numbers for your own hardware.

---

Recommended files

FableFusion-GAIN-F711-27B-IQ4_XS-MTP.gguf
mmproj-FableFusion-GAIN-F711-27B-F16.gguf

The first file is the main model. The second enables image/video input in GGUF runtimes that use an external multimodal projector.

---

llama.cpp quick start

Text

llama-cli \
  -m FableFusion-GAIN-F711-27B-IQ4_XS-MTP.gguf \
  -c 8192 \
  -p "Analyze the trade-offs between speculative decoding and standard autoregressive decoding."

Embedded MTP

llama-cli \
  -m FableFusion-GAIN-F711-27B-IQ4_XS-MTP.gguf \
  --spec-type draft-mtp \
  --spec-draft-n-max 2 \
  -c 8192 \
  -p "Design a robust multi-stage research workflow and justify each stage."

Vision

Use the matching projector:

llama-mtmd-cli \
  -m FableFusion-GAIN-F711-27B-IQ4_XS-MTP.gguf \
  --mmproj mmproj-FableFusion-GAIN-F711-27B-F16.gguf \
  --image image.jpg \
  -p "Describe the image and explain the evidence behind your conclusions."

Use a recent llama.cpp build for Qwen3.8, multimodal and embedded-MTP support.

---

Credits

This model exists because of a long chain of work. Please preserve attribution when mirroring or deriving from it.

• Qwen Team / Alibaba Group — Qwen3.8/Qwen3.x architecture, pretraining, post-training and multimodal foundation.
• Nightmedia / Gheorghe Chesler — F711 source model, final NuSLERP fusion, Architect/Polaris/Fable/F451 branch construction, MXFP4 release, and the benchmark tables quoted above.
• DavidAU — Qwen3.8 Cold-Fusion GAIN V1.1 and multiple upstream Qwen3.6 branches inherited by the Nightmedia lineage.
• armand0e / TeichAI — Qwen3.8 Fable Distill, Fable datasets and Fable-model contributions.
• trohrbaugh — Heretic-base contribution inherited through an upstream DavidAU Qwen3.6 branch.
• Verkiki — local Qwen3.8 IQ4_XS MTP donor used for the embedded MTP block in this release.
• Verkiki — Windows source recovery, MXFP4 → FP16 reconstruction, tensor validation, model-specific imatrix, IQ4_XS release quantization, MTP integration and release packaging.
• ggml-org / llama.cpp — GGUF tooling, IQ4_XS/imatrix tooling, multimodal runtime and MTP runtime support.
• Apple MLX, Hugging Face Safetensors, MergeKit, Unsloth — tooling used directly in the recovery or by the upstream training/merge ecosystem.

Primary upstream links

• Qwen3.8-27B: https://huggingface.co/Qwen/Qwen3.8-27B
• Nightmedia F711 MXFP4 source: https://huggingface.co/nightmedia/Qwen3.8-27B-Fable-Fusion-F711-GAIN-mxfp4-mlx
• DavidAU Cold-Fusion GAIN V1.1: https://huggingface.co/DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1
• TeichAI Qwen3.8 Fable Distill: https://huggingface.co/TeichAI/Qwen3.8-27B-Fable-Distill
• Nightmedia Cold-Fusion GAIN + Fable: https://huggingface.co/nightmedia/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-Fable-mxfp8-mlx
• Nightmedia Qwen3.6 Architect-Polaris2-Fable-B-F451: https://huggingface.co/nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451
• DavidAU Qwen3.6 Heretic2 branch: https://huggingface.co/DavidAU/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking
• DavidAU Qwen3.6 F451 / TRI Polar branch: https://huggingface.co/DavidAU/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic

For redistribution and provenance notes, see NOTICE.md.

---

Benchmark provenance

The benchmark values in this card are transcribed from the upstream model cards linked above.

They should be interpreted as lineage evidence:

• Nightmedia's F711, GAIN, GAIN+Fable and Qwen3.8 baseline values were measured upstream.
• TeichAI's Fable values were measured upstream.
• Qwen's architectural specifications come from the official Qwen3.8-27B model card.
• No claim is made that this exact IQ4_XS binary has reproduced those scores until it is separately benchmarked.

That distinction matters — and it is why the numbers here are useful rather than misleading.