Vision AI For Devs.

Build next-level AI Vision apps with moondream, the tiny open source Vision Language Model (VLM) that runs everywhere and kicks ass.

Get started Try it out

Trusted by the world’s most innovative teams

What can it do?

Answer Questions
Detect Objects
Caption
Count objects BETA Can only count up to about 10 objects so far.

What is that tower?

The tower is the Space Needle, a famous landmark in Seattle, Washington.

What people are saying about Moondream

Brian Roemmele

@BrianRoemmele

More testing of the amazing Moondream open source multimodal LLM today! It is massively small: 1.6B parameter model built using SigLIP, Phi-1.5 and the LLaVA training dataset. I am really impressed. More soon.

MasteringMachines AI

@MstrMachines

Moondream: a 1.6 Billion parameter model that is quite effective and possibly able to go toe to toe with the bigger models in the future.

Luis C

@lucataco93

MoonDream - A tiny vision language model that performs on par w/ models twice its size by @vikhyatk. Its so fast, you might not even catch it streaming output! Try it out now on @replicate

@FieroTy

the 1.6b parameter Moondream model understands it better than GPT4 and Gemini 👀 @vikhyatk

Prasanna Lahoti

@_PrasannaLahoti

Moondream space is amazing, i tried it out yesterday for the very first time after @Suhail tweet

clem 🤗

@ClementDelangue

So impressed with the speed and accuracy of https://huggingface.co/spaces/vikhyatk/moondream1 by @vikhyatk (especially the last answer 😝😝😝). Open multi-modal models have gone a long way!

murat

@mayfer

how is moondream so good? is it because you switched to quality data earlier during pretraining (while plasticity was still high) ?

andrew gao

@itsandrewgao

ever seen an amazing AI image but didn't know how to recreate it? i fine-tuned a #multimodal LLM, Moondream (@vikhyatk), on 35k StableDiffusion image-prompt pairs. 🔗👇 Colab not only describes, but mentions similar artists, styles, mediums [...]