If you're looking for open-source libraries of pre-trained AI models, Cargoship is a good choice. It offers a curated collection of models from HuggingFace and GitHub that you can pull down and use with Docker containers. The service covers a range of tasks, including text processing, text generation, image classification and audio transcription. Cargoship is in early access, but you can host it yourself and keep its models up to date with the latest AI research.
Another good option is LLM Explorer, which catalogs 35,809 open-source Large Language Models. You can filter and compare models based on their attributes, and you can get details on model size, score and memory usage. That makes it a good choice for AI enthusiasts, researchers and industry pros who want to quickly find a good model for their needs.
For generative tasks, Stability AI offers a range of models, including Stable Diffusion 3 for text-to-image generation and Stable LM 2 1.6B for language processing. The service offers flexible membership options and is focused on unbiased and accessible models, so it's a good choice for developers and researchers trying out generative AI.
Replicate is another good option, letting developers run and scale open-source machine learning models. It also includes a library of pre-trained models for a range of tasks and lets you fine-tune with your own data. The service automatically scales and charges by the second, so it's good for everything from generating images to synthesizing speech.