PE-Core-L14-336 is a Perception Encoder model developed by Meta AI. It serves as a state-of-the-art vision encoder for image and video understanding, trained through simple vision-language learning objectives. The model can be used as a CLIP-style model or as a standalone Vision Transformer for feature extraction and zero-shot classification tasks. It is distributed as open weights on Hugging Face for integration with libraries like PerceptionEncoder and timm.
PE Core L14 336 is an Other AI product. It focuses on obtaining high-quality image and video representations without task-specific labeled training data. It is built as an open-source project for machine learning researchers and developers. PE Core L14 336 is open source under the Apache-2.0 license. PE Core L14 336 is available on the web, the command line, and API.
AI at Meta builds and maintains PE Core L14 336, and the product first shipped in 2025. Development happens publicly on GitHub with 2.3k stars. Key capabilities include Zero-Shot Classification, Vision-Language Training, and Image Feature Extraction.
Latest indexed changes and source events
facebook/PE-Core-L14-336 verified by the PulseGate indexer
Other apps tracked under the same category.
