Models 101
What Is a Model?
Description
Discover what an AI model actually is: billions of numbers learned through training, and why running that model, called inference, is the real engineering challenge.
Additional Resources
- Blog: What is AI Inference
- Blog: Inference at Scale
- Product: RDU
Different Types of AI
Description
Get an overview of the main types of AI systems in use today: predictive models, large language models, vision models, and decision-making systems, and learn what each one is designed to do.
Additional Resources
- Blog: Generative AI terms
- Blog: What is an AI chip
What Is an LLM?
Description
Find out how large language models really work: breaking text into tokens, predicting one token at a time, and using transformer attention to decide what comes next.
Related Resources
- Blog: What is AI Inference
- Blog: Inference speed or throughput
- Product: Dataflow Architecture
What Is a Dataflow Graph?
Description
Explore the dataflow graph, the set of nodes and dependencies that maps every calculation an LLM runs, and learn why how that graph executes decides how fast you get an answer.
Related Resources
- Blog: Why dataflow matters
- Blog: Solving AI's infrastructure crisis with dataflow
- Product: Dataflow Architecture
Prefill vs. Decode
Description
Learn how inference splits into two phases, prefill and decode, why one is compute bound and the other memory bound, and how the KV cache makes memory the critical bottleneck.
Related Resources
- Blog: Why dataflow matters
- Blog: Why agentic inference needs hybrid hardware
- Product: RDU
Understanding Prefill & Decode for Disaggregated Inference
Description
Learn how large language models process inputs, what the prefill and decode stages involve, and why disaggregated inference uses different hardware for each to maximize throughput and efficiency.
Additional Resources
- Blog: Why agentic inference needs hybrid hardware
- Blog: Why dataflow matters
- Product: RDU