Overview

SigLIP replaces CLIP’s softmax contrastive loss with a simpler sigmoid loss, which scales better. The So400m version is used as the vision tower of models such as PaliGemma.

Highlights

  • So400m shape-optimized version
  • Zero-shot image classification
  • Image-text retrieval
  • Apache 2.0 license

Typical uses

Zero-shot classification, image search and as an encoder in VLMs.

Things to consider

  • Not a chat model.

Specifications are taken from the model card on Hugging Face and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.