Apple introduced OpenELM, tiny AI language models for smartphones, a proof-of-concept with potential for on-device AI in the future.
OpenELM, though not fully open-source, offers source code availability. It aims to provide efficient language models for local device processing.
Microsoft's Phi-3 models and Apple's OpenELM focus on creating small AI models for local processing. Apple's OpenELM ranges from 270 million to 3 billion parameters.
OpenELM models are categorized into pretrained and instruction-tuned types, each serving different purposes, with a 2048-token context window for training.
[
add
]
[
|
|
...
]