Curio models

Core ML conversions of the search models used by Curio, a macOS app that searches your photos by meaning on-device. Curio downloads these on first launch; you do not need to fetch them by hand.

Folder Model Source Licence
mobileclip-s4/ MobileCLIP-S4 (Apple), Curio's default apple/MobileCLIP-S4 Apple ML Research licence (research use)
clip-vit-b32/ OpenAI CLIP ViT-B/32 openai/clip-vit-base-patch32 MIT

The weights are unchanged apart from conversion to float16 Core ML. Each model's licence still applies to its folder; read Apple's before using MobileCLIP-S4.

Each folder holds image.mlpackage (RGB 0–1 image β†’ unit embedding), text.mlpackage (77 CLIP token ids β†’ unit embedding), the tokenizer's bpe_merges.txt, model.json, and files.json listing every file with its size and SHA-256, which the app verifies after downloading. Produced by scripts/convert-models.sh in the Curio repository.

Downloads last month
46
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support