manuelgscom/factusight-2B-GGUF
Image-Text-to-Text • 2B • Updated • 116
Offline invoice and receipt data extraction with fine-tuned Qwen3.5 vision models (GGUF, llama.cpp). Spanish invoices, JSON output.
Note Flash: fastest, ~2.9 GB VRAM. Low-end GPUs or CPU.
Note Standard: the balanced default, ~4.7 GB VRAM.
Note High Precision: hardest documents, ~7.4 GB VRAM.