MyPrivateClaw

TII Releases Falcon Perception — Open-Weight Multimodal Vision Model for Local Deployment

The Technology Innovation Institute has released Falcon Perception, an open weight vision language model available on Hugging Face. The model supports image un…

Published on MyPrivateClaw

Apr 8, 2026, 8:46 AM UTC

Coverage date

Apr 7, 2026

Last updated

Apr 8, 2026, 8:46 AM UTC

News summary

The Technology Innovation Institute (TII), the Abu Dhabi based research organisation behind the Falcon language model series, has released Falcon Perception , an open weight vision language model (VLM) targeting local deployment on consumer and prosumer hardware. Model Specifications Falcon Perception is available in two sizes: Size Parameters VRAM (Q4 K M) VRAM (FP16) 7B 7 billion 5 GB 14 GB 11B 11 billion 7.5 GB 22 GB The 7B variant runs comfortably on an RTX 3080 (10 GB) or RTX 4070 (12 GB) at Q4 K M quantisation. The 11B variant fits on an RTX 3090 or RTX 4090 at Q4 K M, or requires 24 GB VRAM for FP16 inference. Capabilities Falcon Perception supports: Image understanding : Object recognition, scene description, spatial reasoning Document analysis : OCR, table extraction, form understanding from scanned documents Visual question answering : Multi turn conversations grounded in imag…