Launchpad User Guide
Discover, test, evaluate, and launch curated open-source and enterprise model templates with guided deployment paths.
Who This Guide Is For
- ML engineers
- Application teams
- Platform evaluators
Where To Go
| Page |
Use It For |
/launchpad |
Launchpad overview. |
/launchpad/models |
Browse model catalog. |
/launchpad/playground |
Test candidate models. |
/launchpad/evaluation |
Evaluate selected models. |
/launchpad/monitoring |
Track launch and usage signals. |
Core Concepts
| Concept |
Meaning |
| Model catalog |
A curated list of candidate models and metadata. |
| Template |
A deployment-ready model configuration pattern. |
| Playground |
A test surface for trying prompts and inputs before deployment. |
| Evaluation |
A benchmark or test run to compare candidates. |
| Launch |
The handoff into ModelOps deployment or another managed serving path. |
Common Workflows
Evaluate and launch a model
- Browse model catalog.
- Open a candidate model.
- Test it in the playground.
- Run evaluation if needed.
- Choose a deployment template.
- Send the candidate to ModelOps or the configured launch path.
- Monitor after launch.
Compare candidates
- Filter models by task, license, size, or capability.
- Run the same prompts or eval dataset.
- Compare cost, quality, and deployment requirements.
- Pick the best fit for the target workload.
Best Practices
- Check license, model size, hardware requirements, and data policy before launch.
- Use playground tests only as early screening; run formal evaluations for production.
- Prefer templates that match your security and deployment target.
- Track launched models through ModelOps for ongoing governance.