AITRADER/Jackrong-Qwen3.5-2B-Claude-Reasoning-abliterated-fp16-MLX

Jackrong Qwen3.5-2B Claude Reasoning - Abliterated (bf16 MLX)

This is an abliterated (uncensored) version of Jackrong/Qwen3.5-2B-Claude-4.6-Opus-Reasoning-Distilled, converted to MLX format for Apple Silicon.

What is this model?

The base model is Qwen3.5-2B fine-tuned on Claude 4.6 Opus reasoning data using Unsloth/LoRA, giving it chain-of-thought reasoning capabilities in a compact 2B parameter model. This abliterated version removes safety refusal behaviors while preserving reasoning abilities.

Abliteration Details

• Method: lukey03 (1 direction, norm-preserving, 3 refinement passes, output-only projection)
• Post-processing: LoRA compliance fine-tuning (rank=64, 80 iterations, 32 training examples)
• Format: bf16 MLX (Apple Silicon native)
• Refusal rate: ~14% on 100 adversarial prompts (85% compliance)
• Tool: OBLITERATUS

Architecture

Qwen3.5 hybrid attention architecture:

• 24 layers (6 full attention + 18 linear attention with GatedDeltaNet)
• ~2B parameters
• 262K max context length

Usage

pip install mlx-lm

mlx_lm.generate --model AITRADER/Jackrong-Qwen3.5-2B-Claude-Reasoning-abliterated-fp16-MLX --prompt "Explain quantum computing"

from mlx_lm import load, generate

model, tokenizer = load("AITRADER/Jackrong-Qwen3.5-2B-Claude-Reasoning-abliterated-fp16-MLX")
response = generate(model, tokenizer, prompt="Your prompt here", max_tokens=512)

Quantized Version

A MXFP8 quantized version is available at: AITRADER/Jackrong-Qwen3.5-2B-Claude-Reasoning-abliterated-mxfp8-MLX

Disclaimer

This model is provided for research purposes. Users are responsible for ensuring their use complies with applicable laws and regulations.