Skip to content

EthenEthenEthen

LoginPlatformTry Ethen

Open Source Model Profile · joyfox

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

Qwen3.8-27B-Uncensored-JoyFox-Aggressive is a 27.78B-parameter Qwen image-text-to-text variant from joyfox. Its model card documents low-overrefusal tuning with preserved native vision and MTP components.

Publisher
joyfox
Task
image-text-to-text
Model type
qwen3_5
License
apache-2.0
Library
transformers
Publication status
Accepted · not indexed

Model overview

Qwen3.8-27B-Uncensored-JoyFox-Aggressive is published by joyfox as an image-text-to-text model. The captured configuration identifies Qwen3_5ForConditionalGeneration, and Safetensors metadata reports 27,781,427,952 parameters (about 27.78B). According to the model card, it is a direct-response, low-overrefusal variant of Qwen/Qwen3.8-27B that keeps the native tokenizer, vision stack, and MTP components intact.

Recorded capabilities

Low-overrefusal response style

According to the model card, this release reduces false refusals and policy-style preambles while retaining the native architecture and general capabilities of Qwen/Qwen3.8-27B.

Preserved native stack

The model card says the checkpoint preserves the official tokenizer, chat template, configuration, vision stack, multimodal projector, and native Multi-Token Prediction weights.

GGUF and llama.cpp workflow

Hub tags include gguf, and the card documents llama.cpp usage, a Q6_K GGUF file, and a matching F16 mmproj for multimodal inference.

Use cases in the source record

  • Direct-response conversational workflows in English or Chinese that use the card's documented low-overrefusal behavior.
  • Multimodal experiments that pair the Q6_K GGUF file with its matching F16 mmproj following the card's llama-cli example.

Limitations and unknowns

  • No evaluation results were extracted from this record.
  • No context-window value was extracted from this record; the card's 32768 figure appears only inside a llama-cli example command.
  • Provider state is historical snapshot data, not independently refreshed current availability.
  • Behavioral and compatibility claims come from the publisher model card and have not been independently verified by Ethen.

Source and provenance

Source: joyfox/Qwen3.8-27B-Uncensored-JoyFox-Aggressive

Captured: Unknown. Processed: 2026-09-07T19:34:48.578928+00:00.

🦊 Qwen3.8-27B-Uncensored-JoyFox-Aggressive A direct-response, low-overrefusal variant of Qwen3.8-27B with native model components preserved. English · 简体中文 🦊 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 sepa…

F001F002F003F004F005F006F007F009F010F011F012F013F014F015F016F017F018F019F020