CoolFace
Modelpublic

PocketWeights/PocketWeights-Qwen2.5-14B-Coder-Creative-GGUF

sourceHugging Faceapache-2.0updated 27d agoView on Hugging Face
0likes243downloads
Model Card

⚡ PocketWeights: Qwen2.5 14B Coder-Creative (GGUF)

Heavy models, made light. PocketWeights optimizes state-of-the-art open-source LLMs into efficient GGUF formats so you can run unrestricted, powerful AI locally on consumer hardware, gaming laptops, and edge devices.

🧠 About This Model

These are the official, first-party GGUF quantizations of [PocketWeights-Qwen2.5-14B-Coder-Creative](https://huggingface.co/PocketWeights/PocketWeights-Qwen2.5-14B-Coder-Creative).

This model is a cross-domain synthesis engineered to solve a common trade-off in mid-sized language models. Using the DARE-TIES algorithm, we injected the high-density coding logic of Qwen2.5-Coder-14B directly into an uncensored, extended-context Qwen2.5-14B-abliterated foundation.

🎯 Key Highlights

  • —SOTA Syntax & Logic: Top-tier Python, C++, Rust, and shell script capabilities inherited from the Qwen 2.5 Coder architecture.
  • —Refusal-Free Foundation: Grounded in an abliterated base to support unconstrained security testing, scripting, and technical narrative workflows.
  • —The 14B Sweet Spot: Optimal performance-to-compute ratio—small enough to run smoothly on 12GB–16GB VRAM hardware while outperforming 7B/8B models in architectural depth.
⚠️ Disclaimer: This model has had its corporate safety guardrails removed. It is designed for researchers, writers, and developers operating in secure, local environments.

📦 Available Files & Hardware Requirements

We provide highly curated, precision-focused files—no clutter, just the formats you actually need.

Quant FormatFile SizeRecommended HardwareBest For
Q4_K_M~8.9 GB12 GB VRAM / 16 GB RAMThe Balanced Standard. The community sweet spot offering maximum speed and high intelligence for RTX 3060, RTX 4070, and 16GB MacBooks.
Q6_K~12.1 GB16 GB VRAM / 24 GB RAMNear-Lossless Quality. Maximizes coding syntax accuracy and vocabulary richness for RTX 4080 and high-end desktop rigs.
Q8_0~15.7 GB24 GB VRAM / 32 GB RAMEnterprise Fidelity. Practically indistinguishable from the base model. Ideal for RTX 3090/4090 or Apple Silicon (M1/M2/M3 Max).

🚀 Quick Start Guide

You can run this model offline on your local machine in under 2 minutes:

Option 1: LM Studio (Visual GUI — Easiest)

  1. 1.Download and install [LM Studio](https://lmstudio.ai/) (Free for Windows, macOS, and Linux).
  2. 2.Click the Magnifying Glass (Search) icon in the left navigation bar.
  3. 3.Search for: PocketWeights/PocketWeights-Qwen2.5-14B-Coder-Creative-GGUF
  4. 4.Click Download next to your preferred size (Q4KM is recommended), open the Chat Tab, load the model at the top, and start coding!

Option 2: Ollama (Terminal / CLI)

Run the balanced tier immediately from your terminal (Ollama will automatically pull the Q4KM):

bash
ollama run hf.co/PocketWeights/PocketWeights-Qwen2.5-14B-Coder-Creative-GGUF

Option 3: llama.cpp

Execute directly via llama.cpp, offloading maximum layers to your GPU:

bash
llama-cli -m PocketWeights-14B-Coder-Creative-Q4_K_M.gguf -p "Write a Python script using scapy to analyze packet headers." -ngl 40

🤝 Support the PocketWeights Mission

I build, verify, and maintain these quantization pipelines to provide high-quality, unrestricted, and hardware-friendly models to the open-source community for free.

Running conversion setups, cloud instances, and storage requires ongoing resources. If these weights have saved you time, compute overhead, or API bills, please consider supporting the project with a small tip!

☕ Donation Options

Ko-fi: ko-fi.com/iamvishalnarayan

Web3 / Crypto (Polygon / ETH):

0x4FC189bf839A89259dd28DE8cD97883c49e15615

Tip: Sending via the Polygon network keeps transfer gas fees below $0.01!

📄 Attribution & License

Synthesis Lab: PocketWeights

Base Architecture: Alibaba Cloud (Qwen2.5)

Abliteration Source: qq591503 / huihui-ai

License: Apache-2.0