language:

• en
  library_name: mlc-llm
  base_model: huihui-ai/Huihui-Qwen3.5-2B-abliterated
  license: other
  tags:
• mlc-llm
• web-llm
• webgpu
• qwen3.5
• abliterated
• uncensored
• q4f16_1
  pipeline_tag: text-generation

Huihui Qwen3.5-2B (abliterated) — MLC q4f16_1

MLC-converted (q4f16_1) build of the abliterated
huihui-ai/Huihui-Qwen3.5-2B-abliterated
model, for running fully client-side in the browser via
WebLLM (WebGPU).

• Base model: huihui-ai/Huihui-Qwen3.5-2B-abliterated (all credit to Huihui for the abliteration)
• Quantization: q4f16_1
• Architecture: Qwen3.5-2B (model_type: qwen3_5, hidden 2048, 24 layers)
• Converted with: mlc-llm convert_weight (source build)

Usage (WebLLM)

import * as webllm from "https://esm.run/@mlc-ai/[email protected]";

const appConfig = {
  model_list: [
    {
      model: "https://huggingface.co/RaiseRuntimeError/Huihui-Qwen3.5-2B-abliterated-q4f16_1-MLC",
      model_id: "Huihui-Qwen3.5-2B-abliterated-q4f16_1-MLC",
      model_lib:
        webllm.modelLibURLPrefix + webllm.modelVersion + "/Qwen3.5-2B-q4f16_1_cs1k-webgpu.wasm",
      vram_required_MB: 2245.44,
      overrides: { context_window_size: 4096, max_history_size: 1 },
    },
  ],
};

const engine = await webllm.CreateMLCEngine(
  "Huihui-Qwen3.5-2B-abliterated-q4f16_1-MLC",
  { appConfig },
);

Notes

This is a derivative of a third-party abliterated model, re-quantized to MLC
q4f16_1 for browser inference. The abliteration itself is the work of
huihui-ai.