Downloads · 30 days
46
10% of all-time downloads
PocketWeights/Gliese-Qwen3.5-9B-Abliterated-WebGGUF
Gliese-Qwen3.5-9B-Abliterated-WebGGUF is a image-text-to-text model from PocketWeights. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for gguf. The card lists the license as apache-2.0.
Heavy models, made light. PocketWeights optimizes state-of-the-art vision and language models into specialized GGUF formats designed for browser deployment, edge devices, and local consumer hardware.
Downloads · 30 days
46
10% of all-time downloads
All-time downloads
463
Public
Repo size
24.2 GB
Likes
1
Public
Click a slice to open those files.
.gguf24.2 GB · 100%
From the Hugging Face model README
Heavy models, made light. PocketWeights optimizes state-of-the-art vision and language models into specialized GGUF formats designed for browser deployment, edge devices, and local consumer hardware.
Gliese-Qwen3.5-9B-Abliterated-Caption (by prithivMLmods) is an advanced 9-billion parameter Vision-Language Model (VLM) specialized in high-detail image captioning, visual question answering (VQA), and deep optical scene understanding.
| Quant Format | File Size | Recommended Hardware | Best Use Case |
|---|---|---|---|
| Q4_0 | 4.95 GB | 6 GB VRAM / 8 GB RAM | WebLLM Standard. Lowest memory footprint and standard baseline for browser WebGPU deployments. |
| Q4_1 | 5.41 GB | 6 GB VRAM / 8 GB RAM | Enhanced 4-Bit. Slightly higher precision scaling for fine-grained image detail extraction. |
| Q5_0 | 5.87 GB | 8 GB VRAM / 12 GB RAM | High Precision Standard. Excellent balance of visual fidelity and legacy CPU execution speed. |
| Q5_1 | 6.33 GB | 8 GB VRAM / 16 GB RAM | Maximum Visual Fidelity. Captures subtle color grading, complex text/OCR, and dense visual layouts. |
You can run this vision model locally or integrate it directly into web apps:
Launch the model immediately in your terminal:
# Run the standard 4-bit WebGGUF
ollama run hf.co/PocketWeights/Gliese-Qwen3.5-9B-Abliterated-WebGGUF:Q4_0
# Or run the high-precision 5-bit version
ollama run hf.co/PocketWeights/Gliese-Qwen3.5-9B-Abliterated-WebGGUF:Q5_0
Download and open LM Studio (Free for Windows, macOS, and Linux).
Click the Search icon on the left bar and search for:
PocketWeights/Gliese-Qwen3.5-9B-Abliterated-WebGGUF
Download Q4_0 or Q5_0, head to the Chat Tab, load the model, and attach images directly into the chat!
These weights are pre-formatted for direct integration into client-side WebGPU projects:
import * as webllm from "@mlc-ai/web-llm";
// Load Q4_0 directly into browser memory
const engine = await webllm.CreateMLCEngine(
"PocketWeights/Gliese-Qwen3.5-9B-Abliterated-WebGGUF",
{ initProgressCallback: console.log }
);
Download your desired .gguf file and matching mmproj vision projector file from the Files and versions tab above.
Execute multimodal image analysis via CLI:
llama-minicpmv-cli -m Gliese-Qwen3.5-9B-Abliterated-WebGGUF-Q4_0.gguf --mmproj mmproj-model-f16.gguf --image test_image.jpg -p "Describe this image in extreme detail."
I build, verify, and maintain automated pipelines to deliver lightweight, unrestricted, and hardware-friendly models to the open-source community completely for free.
Running conversion clusters, web hosting, and continuous build pipelines requires ongoing compute resources. If these models have saved you time, compute overhead, or server hosting bills, please consider supporting the project with a small tip!
Buy me a coffee on Ko-fi: ko-fi.com/iamvishalnarayan
Web3 / Crypto (Polygon / ETH):
0x4FC189bf839A89259dd28DE8cD97883c49e15615
Tip: Sending via the Polygon network keeps network gas fees below $0.01!
Base Architecture: Qwen Team / Alibaba Cloud
Finetune & Abliteration: Created by prithivMLmods
License: Apache 2.0 (Permissive open-source license for personal and commercial applications)