HuggingSnap

Earlier this week, the folks at Hugging Face released HuggingSnap, a new on-device “visual intelligence” app for iPhone. The app is very similar to Apple’s own Visual Intelligence feature for iOS 18, but what makes this stand out is that it uses the smolvlm2 vision model (specifically, the 500M one built with Swift MLX) to process everything locally without sending any data to the cloud. You can simply point your camera at something, ask questions about it, and get responses in real-time. In my tests, it worked well with scenes around me, but it struggled a little when it came to text recognition. Regardless, it’s another cool example of local, open-source AI being comparable to cloud-first models.