Jarvis (Qwen/Qwen3.8-27B - HSLQ-v2 Compressed)
This repository contains the Jarvis-HSLQ-v2 compressed checkpoint of Qwen/Qwen3.8-27B.
Compression & Fidelity Telemetry
- Original Model Size (BF16): ~50.3 GB
- Jarvis Compressed Physical Size:
8.01 GB (`2.23 bits/weight`) - Layer Output Cosine Similarity:
99.3084%
Quick Dequantization Usage
Import dequantize_jarvis_tensor from the included jarvis_loader.py file in this repository.
- Downloads last month
- 119
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for shivrajgurjar/jarvis
Base model
Qwen/Qwen3.8-27B