joyfox

Qwen3.8-27B-Uncensored-JoyFox-Aggressive

README

License: apache-2.0

🦊 Overview

This release focuses on reducing false refusals and unnecessary policy-style preambles while retaining the native architecture and general capabilities of Qwen/Qwen3.8-27B.

The final checkpoint preserves the official tokenizer, chat template, and configuration, the complete vision stack and multimodal projector, native Multi-Token Prediction (MTP) weights. Vision tensors remain unchanged and are verified separately. MTP tensors remain unchanged and are verified separately.

⚡ What “Aggressive” means

“Aggressive” describes the strength of the low-overrefusal behavior. The intended response style is direct and substantive for benign, fictional, analytical, educational, authorized, privacy-respecting, or otherwise legitimate requests, including requests whose wording is sensitive or ambiguous.

The intervention does not add factual knowledge. The model can still hallucinate, accept a false premise, misunderstand intent, or produce unsuitable content. Users should evaluate it for their own application and remain responsible for deployment controls and applicable law.

🧬 Architecture and behavior

The repository contains a complete standalone checkpoint in the native Qwen3.8 tensor layout. Its text weights favor direct, substantive answers to legitimate requests, while the tokenizer, chat template, configuration, vision stack, multimodal projector, and MTP components retain upstream compatibility.

Release validation enforces these invariants:

official tokenizer, chat template, configuration, and special-token IDs are unchanged;
vision encoder and projector tensor names, shapes, dtypes, and values match the official checkpoint;
native MTP tensor names, shapes, dtypes, and values match the official checkpoint before conversion;
the public native checkpoint contains only the complete model and inference metadata;
BF16, GGUF, MTP, and multimodal inference are tested independently.

🧪 Evaluation

The reported evaluation contains 1,000 prompts spanning difficult legitimate requests, bilingual exact-answer controls, and a non-blocking behavior diagnostic.

Table with columns: Split, Prompts, Purpose
Split	Prompts	Purpose
Benign overrefusal	800	difficult legitimate and ambiguity-sensitive requests
Bilingual capability controls	100	deterministic Chinese/English exact-answer checks
Non-blocking behavior diagnostic	100	descriptive V1 reporting

Table with columns: Checkpoint, Benign refusal rate, Capability score
Checkpoint	Benign refusal rate	Capability score
Official Qwen3.8-27B	59.2%	59.0%
JoyFox BF16	0.0%	59.0%

All results are generation-based. The fixed suite contains 1,000 prompts, greedy decoding is used, and the tested llama.cpp runtime commit is 885c5bbe8e04.

💾 GGUF downloads

All main GGUF files are generated with an importance matrix. Native MTP tensors are bundled in every main GGUF, and a standalone MTP GGUF is also released for split-draft runtimes. The vision stack is exported as an F16 mmproj and validated with real image input.

Table with columns: File, Quant, Intended use, Size
File	Quant	Intended use	Size
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q2_K.gguf`	Q2_K	minimum footprint	10.12 GiB
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q3_K_M.gguf`	Q3_K_M	low-memory deployment	12.57 GiB
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q4_K_M.gguf`	Q4_K_M	recommended 4-bit balance	15.66 GiB

Compatibility builds without MTP

The following -no-mtp files contain the same target-model trunk but no MTP/NextN draft tensors. Use one of these when a runtime version cannot load a GGUF with bundled MTP. They support normal text generation and can use the same matching mmproj; speculative MTP decoding is intentionally unavailable. A runtime must still support the Qwen3.8/Qwen3.5 GGUF architecture itself.

Table with columns: File, Quant, Size
File	Quant	Size
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q2_K-no-mtp.gguf`	Q2_K	9.98 GiB
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q3_K_M-no-mtp.gguf`	Q3_K_M	12.39 GiB
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q4_K_M-no-mtp.gguf`	Q4_K_M	15.41 GiB
`Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q5_K_M-no-mtp.gguf`	Q5_K_M	17.91 GiB

All six compatibility builds were inspected as 851-tensor target models with zero MTP/NextN tensors and passed a real text-generation smoke test.

llama.cpp has no standard Q7 K-quant, so the practical Q2–Q8 matrix uses Q2, Q3, Q4, Q5, Q6, and Q8.

🚀 Usage

Use llama.cpp commit 885c5bbe8e04 or a newer compatible build.

💬 Text chat

bash
llama-cli \
  -m Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q5_K_M.gguf \
  --jinja -c 32768 -ngl 99

⚡ Native bundled MTP

bash
llama-server \
  -m Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q6_K.gguf \
  --jinja -c 32768 -ngl 99 \
  --spec-type draft-mtp --spec-draft-n-max 3

MTP is an optional speculative-decoding feature. Its benefit depends on draft acceptance, context length, backend, and available memory.

👁️ Multimodal inference

bash
llama-cli \
  -m Qwen3.8-27B-Uncensored-JoyFox-Aggressive-Q6_K.gguf \
  --mmproj mmproj-Qwen3.8-27B-Uncensored-JoyFox-Aggressive-F16.gguf \
  --image example.png \
  --prompt "Describe this image accurately." \
  --jinja -c 32768 -ngl 99

The mmproj must match this checkpoint. It is kept at F16 to preserve visual quality.

🎛️ Recommended settings

Use the official chat template and upstream sampling defaults. For reproducible evaluation, the scores above use deterministic greedy decoding.

⚠️ Validation and limitations

Before upload, the release must pass native checkpoint inspection, component-preservation checks, every-quant text generation, bundled and standalone MTP generation, and real-image generation with the matching F16 mmproj. The uploaded repository is then checked against the local release inventory.

Lower-bit quantization can reduce factual precision, multilingual consistency, long-context stability, visual grounding, and subtle instruction following. Abliteration can also weaken refusal behavior more broadly than intended. Evaluate the chosen quant and deployment policy for the actual use case.

📜 License and attribution

Released under Apache-2.0, following the base-model license. Qwen is created by the Qwen team. GGUF conversion and quantization use llama.cpp.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Model Details

Model Provider