Q3.5-27B-Opus-DA

Q3.5-27B-Opus-DA (Qwen3.5 Distilled-Abliterated) is a reasoning-focused model built on Qwen/Qwen3.5-27B. This model is optimized for rich, detailed, and context-aware reasoning, leveraging multi-stage training on Opus reasoning traces combined with advanced refusal direction analysis and ablation-based training strategies to reduce internal refusal behaviors while preserving strong reasoning and instruction-following performance.

This model is intended strictly for research and learning purposes. Due to reduced internal refusal mechanisms, it may generate sensitive or unrestricted content. Users assume full responsibility for how the model is used. The authors and hosting platform disclaim any liability for generated outputs.

Note: This model is experimental and may generate artifacts.

Key Highlights

• Multi-Stage Training on Opus Reasoning Traces: Fine-tuned across multiple stages using high-quality reasoning traces from Claude Opus, enabling deep reasoning alignment.
• Distilled Abliterated (DA): Applies advanced refusal direction analysis and ablation-based strategies to reduce internal refusal behaviors while preserving reasoning quality.
• Qwen3.5 Base: Built on the powerful Qwen/Qwen3.5-27B backbone for strong reasoning and generation performance.
• Instruction + Reasoning Fusion: Handles both instruction-following and complex multi-step reasoning tasks seamlessly.
• High-Coherence Outputs: Maintains consistency across long generations with improved contextual grounding.

Datasets Used and Training Details

Category	Details
Base Model	Qwen/Qwen3.5-27B
Final Model Size	27B Parameters
Training Type	Multi-stage distillation + abliteration
Objective	Preserve reasoning quality from larger models; reduce refusal behaviors via ablation strategies; improve instruction-following reliability
Reasoning Dataset 1	nohurry/Opus-4.6-Reasoning-3000x-filtered
Reasoning Dataset 2	Jackrong/Qwen3.5-reasoning-700x
Reasoning Dataset 3	Roman1111111/claude-opus-4.6-10000x
Alignment / Evaluation Dataset	prithivMLmods/harm_bench
Training Focus	Structured reasoning, long-chain thinking, robustness across diverse prompts
Training Inspiration 1	Jackrong-llm-finetuning-guide
Training Inspiration 2	Unsloth Documentation

Quick Start with Transformers

pip install transformers==5.7.0
# or latest
pip install git+https://github.com/huggingface/transformers.git

from transformers import Qwen3_5ForConditionalGeneration, AutoProcessor
import torch

model = Qwen3_5ForConditionalGeneration.from_pretrained(
    "prithivMLmods/Q3.5-27B-Opus-DA",
    torch_dtype="auto",
    device_map="auto"
)

processor = AutoProcessor.from_pretrained(
    "prithivMLmods/Q3.5-27B-Opus-DA"
)

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Generate a highly detailed caption of a futuristic city skyline at sunset."}
        ],
    }
]

text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)

inputs = processor(
    text=[text],
    padding=True,
    return_tensors="pt"
).to("cuda")

generated_ids = model.generate(**inputs, max_new_tokens=512)

generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]

output_text = processor.batch_decode(
    generated_ids_trimmed,
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False
)

print(output_text)

Intended Use

• Reasoning & Chain-of-Thought Tasks: Deep multi-step reasoning powered by Opus-distilled traces
• Instruction Following: Hybrid prompts requiring both instruction adherence and reasoning
• Red-Teaming & Alignment Research: Evaluating reduced-refusal systems and refusal direction analysis
• Local High-Performance Deployment: Multi-GPU or quantized inference setups
• Research on Abliteration: Studying the effects of ablation-based training on reasoning preservation

Limitations & Risks

Important Note: This model intentionally minimizes built-in safety refusals.

• Sensitive Content Risk: May produce unrestricted or controversial outputs
• User Responsibility: Requires careful and ethical usage
• High Compute Demand: Large models need significant VRAM or optimized inference
• Abliteration Trade-offs: Reduced refusal may impact safety alignment and output filtering