Winnougan
PulseAugur coverage of Winnougan — every cluster mentioning Winnougan across labs, papers, and developer communities, ranked by signal.
-
Qwen3.5 models quantized for ComfyUI, run on 8GB VRAM
A user has released quantized versions of the Qwen3.5 models (2B, 4B, and 9B parameters) for use with ComfyUI. These INT8 models, utilizing ConvRot quantization, are designed to run efficiently on systems with as little…
-
ComfyUI gains runtime quantization for faster AI image generation
A new ComfyUI node called QuantFunc has been developed to enable runtime 4-bit quantization of AI models, significantly speeding up inference times. This allows users to apply quantization on-the-fly without needing pre…
-
Krea 2 AI models released in quantized formats for NVIDIA GPUs
Winnougan has released quantized versions of the Krea 2 Base and Krea 2 Turbo models, optimized for various NVIDIA GPU architectures. These versions utilize formats like NVFP4, FP8, MXFP8, and INT8, with specific recomm…