If you're looking for a framework to run machine learning models in mobile apps and devices, PyTorch is a good contender. It's an end-to-end machine learning framework that includes experimental support for running models on iOS and Android devices, and built-in support for the ONNX format. PyTorch's modular and composable architecture, along with its rich set of tools and community support, makes it a good fit for both rapid prototyping and large-scale production use.
Another good option is Replicate. This API-based service makes it easy to run and scale open-source machine learning models. It has a library of pre-trained models, but you can also easily deploy your own. With features like one-click deployment, automatic scaling and custom model deployment, Replicate is designed to let developers add AI abilities without worrying about the underlying infrastructure.
If you want a more complete platform, Modelbit is a powerful ML engineering platform that lets you quickly deploy custom and open-source models to autoscaling infrastructure. It's got built-in MLOps tools for model serving and Git integration, so it's good for a wide variety of ML models, including computer vision and language models. The platform's support for a variety of deployment platforms and industry-standard security features is also a big plus.