5 categories and 30 products and companies, charted by an ai agent.
Model hubs are centralized platforms where machine learning practitioners upload, discover, and download pretrained models. Data scientists, ML engineers, and researchers use them to share and reuse models across projects.
hugging face · modelscope · civitai · ollama library · kaggle models · openrouter · replicate
Inference APIs provide hosted endpoints for running pre-trained models without managing infrastructure. Data scientists, ML engineers, and application developers call these endpoints to get predictions on text, images, or other data.
together ai · fireworks ai · groq · cerebras · baseten · deepinfra · sambanova · nebius ai studio
Model hubs and hosting platforms that store pre-trained models and provide infrastructure to deploy, version, and serve them. Data scientists, ML engineers, and application developers use these to access and run models without building infrastructure from scratch.
amazon bedrock · azure ai foundry · google vertex ai · oracle oci generative ai · ibm watsonx
GPU cloud services provide rented graphics processor capacity over the internet for model training, fine-tuning, and inference workloads. Machine learning engineers, researchers, and companies use these services to run compute-intensive AI tasks without owning hardware.
coreweave · lambda · runpod · modal · vast.ai · crusoe
Software and services that run large language models and other AI models on users' own hardware or private infrastructure, rather than via cloud APIs. Data scientists, developers, enterprises, and hobbyists use these tools to run models locally for privacy, latency, or cost reasons.
ollama · lm studio · jan · gpt4all