SenseNova-U1-8B-MoT-Infographic-V2 (int8_convrot + bf16)
SenseNova's SenseNova-U1-8B-MoT-Infographic-V2, quantized down so it fits in ComfyUI without renting a GPU cluster.
It's a diffusion model built for infographics — dense, text-heavy poster-style images where the little words are supposed to stay readable.
What's in here
| File | Precision | Size |
|---|---|---|
SenseNova-U1-8B-MoT-Infographic-V2.bf16.safetensors | bf16 | ~33 GB |
SenseNova-U1-8B-MoT-Infographic-V2.int8_convrot.safetensors | int8_convrot | ~18 GB |
Grab the int8_convrot one if you want to save VRAM. Grab the bf16 one if you've got the memory to spare and want the full-fat original weights.
How the int8 quant was made
The int8_convrot file was quantized with Comfy-Org/comfy-model-tools using its int8_convrot scheme. The bf16 file is just the original weights repacked into a single safetensors.
Credit
The model itself is all SenseNova's work — I just shrank the files. Go read their model card for the real details. Apache 2.0, same as the original.