GraySoft
Projects Models Compare Cloud benchmarks FAQ Download guIDE →
Model Intelligence Sheet

OpenIntelligenceNet/Spark-X2.5-1.7B-Fable5.1-Distilled-Zero-Refusal-GGUF overview

Heretic Spark X2.5 1.7B GGUF Model Description This repository contains the GGUF formats FP16 and Q4 K M for the uncensored Spark X2.5 1.7B architecture. The m…

ggufuncensoredimatrixquantizationreasoningendpoints_compatibleregion:usconversational

Runs locally from ~1.03 GB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).

Downloads
0
Likes
0
Pipeline

Repository Files & Downloads

2 GGUF files detected
Direct downloads for local inference
FileTypeQuantizationSizeLink
Spark-1.7B-F16.ggufGGUFF163.19 GBDownload
Spark-1.7B-Q4_K_M-Imatrix.ggufGGUFQ4_K_M1.03 GBDownload

Model Details

Model IDOpenIntelligenceNet/Spark-X2.5-1.7B-Fable5.1-Distilled-Zero-Refusal-GGUF
AuthorOpenIntelligenceNet
Pipeline
License
Base model
Last modified2026-09-19T14:27:04.000Z

Model README

---

tags:

  • gguf
  • uncensored
  • imatrix
  • quantization
  • reasoning

---

Heretic-Spark-X2.5-1.7B-GGUF

Model Description

This repository contains the GGUF formats (FP16 and Q4_K_M) for the uncensored Spark-X2.5 1.7B architecture. The model is stripped of standard alignment restrictions to allow direct and unbounded responses to complex, creative, and reasoning-based queries.

Quantization & Imatrix Calibration

The Q4_K_M quantization was driven by a robust Importance Matrix (imatrix) to preserve structural integrity. The matrix was calibrated using over 2.14 million high-quality tokens from deep reasoning traces, advanced mathematics, uncensored instruction logic, and conversational roleplay.

Available Files

  • Spark-1.7B-F16.gguf: Uncompressed 16-bit precision base file for maximum accuracy.
  • Spark-1.7B-Q4_K_M-Imatrix.gguf: Highly efficient 4-bit quantization, calibrated via custom imatrix for an optimal balance of speed, VRAM usage, and reasoning fidelity.

Run OpenIntelligenceNet/Spark-X2.5-1.7B-Fable5.1-Distilled-Zero-Refusal-GGUF with guIDE

Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.

Download guIDE → · Browse 524k+ models · Compare models

Source: Hugging Face · Compare models