Siglip 2 Giant Opt Patch 16 384
Google's Siglip 2 Giant Opt Patch 16 384 is a vision model suited for tasks such as zero-shot image classification and image-text retrieval. It is genuinely best at semantic understanding, localization, and dense features, thanks to its unified pretraining objective that combines prior techniques.
Input
Output
Context
0K
Max Output
-
Parameters
1.9B
Technical Specifications
Model TypeVision
Context Window256 tokens
Max Output TokensNot available
Parameters1.9B
Release DateFeb 17, 2025
Training CutoffFeb 20, 2025
Licenseapache-2.0
Open SourceYes
Input Modalities
Image
Output Modalities
Text
Capabilities
Estimated GPU Requirements for Siglip 2 Giant Opt Patch 16 384
Estimates based on INT8 quantization. Actual requirements vary by framework and configuration.
Related Tools
Data sourced from official provider APIs and documentation
Last updated: Jun 24, 2026
Start building with the right model.
Automatically route workloads to the right model for every task, every time.