Model Architectures · Practitioner
Model Garden and Choosing a Model on Evidence
Hundreds of models — Gemini, Imagen, Veo, Gemma, Claude, Llama and more — behind one platform. Learn the deployment shapes and how to choose on cost, latency and capability rather than benchmark noise.
- 1.1 hours lesson
- intermediate level
- Google Cloud AI Track
- Narrated presentation
What you’ll learn
- Model Garden as catalogue plus deployment surface across first-party, partner and open models
- Three deployment shapes: managed API, self-deployed endpoint, notebook/tuning
- Self-deployed endpoints bill per hour while idle — the platform's biggest cost trap
- Choosing on a scored sample of your own task, not public leaderboards
- Regional model availability and partner-model enablement steps
- Pinning model versions and tracking deprecation schedules
Inside this lesson
A narrated, slide-by-slide presentation with 1 prediction exercise, 1 knowledge check, 4 videos. You answer as you go, and the platform checks your work.