Model `ornith-1.0-9b-mtp` wrongly reported as "Size: 71.2 GB · VRAM: 71.2 GB" (size of the entire HF repository) in "Install Models" gallery

[ackalker]

LocalAI version:
v4.6.1, installed using Linux launcher

Environment, CPU architecture, OS, and Version:
Linux miki-linux-ai 7.1.3-1-cachyos #1 SMP PREEMPT_DYNAMIC Sat, 04 Jul 2026 20:05:25 +0000 x86_64 GNU/Linux
CPU: Ryzen 9950X3D
RAM: 64 GB
GPU: RTX 5080

Describe the bug
In the "Install Models" gallery, the model ornith-1.0-9b-mtp is wrongly reported as "Size: 71.2 GB · VRAM: 71.2 GB" and "May not fit", which is to be expected for my RTX 5080 GPU, as the reported size is the size of the entire Hugging Face repository, not just the needed GGUF file.

To Reproduce

1. In the LocalAI Web UI, switch to "Import Models" tab.
2. Search for "ornith".

Expected behavior
The model's actual size as per the model file on Hugging Face should be reported, not the size of the entire repository, which contains several model variants with different quantizations.

Logs

Additional context
The model definition in gallery/index.yaml correctly lists the model file URI as: https://huggingface.co/protoLabsAI/Ornith-1.0-9B-MTP-GGUF/resolve/main/ornith-9b-mtp-kl-Q4_K_M.gguf

[ackalker]

The above URI returns a 404. the correct URI should be: https://huggingface.co/protoLabsAI/Ornith-1.0-9B-MTP-GGUF/resolve/main/Ornith-1.0-9B-MTP-Q4_K_M.gguf

This is because the files in the Hugging Face repository have been reorganized, see: https://huggingface.co/protoLabsAI/Ornith-1.0-9B-MTP-GGUF/commit/7bcb8f00ba623e82b1659fcf28aee3e4125307bf

Please update the model into the gallery, using the correct URI.