Hugging Face
Underdog Saluki 27B (Under 8 GB)
Qwen3.8-27B parameter model optimized to run locally in under 8 GB of VRAM.
Quantized
Engineered by Conway Research, Underdog-Saluki brings the full capability of 27B parameter intelligence onto consumer laptops and budget GPUs.
Fits in under 8 GB of VRAM while preserving benchmark reasoning and prompt compliance.
What's inside
- 01Runs in under 8 GB VRAM
- 02Low-bit precision quantization
- 03Local Ollama / vLLM compatible
Keep exploring.
All resourcesModelsOpen Source
MoCha: Video Character Replacement
End-to-end video character replacement framework with realistic motion preservation.
AI ModelView
ModelsOpen Weights
Qwen3.8-27B — OBLITERATED
Uncensored and optimized 27B parameter open-weights reasoning model.
Hugging FaceView
ModelsVideo AI
FreeVideo: Open Video Generation
Free open-source Higgsfield alternative for high-fidelity text-to-video generation.
AI ModelView