---

base_model: coder3101/Qwen3.5-2B-heretic

base_model_relation: quantized

library_name: gguf

license: apache-2.0

language:

• multilingual

pipeline_tag: image-text-to-text

tags:

• gguf
• text-generation
• multimodal
• vision
• reasoning
• thinking
• agentic
• tool-use
• abliterated
• uncensored
• qwen3
• imatrix
• conversational

---

Qwen3.5-2B-heretic-imatrix — GGUF

> [!NOTE]

> This card is a work in progress. Full architecture details, abliteration parameters, and inference instructions will be added shortly.

This repository hosts importance-matrix (imatrix) calibrated GGUF weights and the associated vision projection matrix for Qwen3.5-2B-heretic, quantized from the source floating-point tensors provided by coder3101/Qwen3.5-2B-heretic.

Repository Files

| Filename | Format | Size | llama.cpp Build | Description |

|---|---|---|---|---|

| Qwen3.5-2B-heretic-IQ4_NL-imatrix.gguf | IQ4_NL | 1.23 GB | b9860 | Main model weights, imatrix-calibrated, smaller non-linear 4-bit quant |

| Qwen3.5-2B-heretic-Q4_K_M-imatrix.gguf | Q4_K_M | 1.27 GB | b9860 | Main model weights, imatrix-calibrated |

| Qwen3.5-2B-heretic-Q5_K_M-imatrix.gguf | Q5_K_M | 1.41 GB | b9860 | Main model weights, imatrix-calibrated, higher fidelity |

| mmproj-Qwen3.5-2B-heretic-BF16.gguf | BF16 | 671 MB | b9860 | Vision projector (full precision) |

| mmproj-Qwen3.5-2B-heretic-Q8_0.gguf | Q8_0 | 365 MB | b9860 | Vision projector (compact) |

| mtp-Qwen3.5-2B-heretic-BF16.gguf | BF16 | 1.03 GB | b9870 | Multi-token prediction head (full precision) |

| mtp-Qwen3.5-2B-heretic-Q8_0.gguf | Q8_0 | 551 MB | b9870 | Multi-token prediction head (compact) |