Recipe: pick a model for a task

Recipe — pick the right model for a task. Pointer to the canonical SDK quickstart + filtering sample.

The canonical pattern lives in the SDK:

• samples/core/quickstart.py — your first end-to-end evaluation in <30 lines
• samples/core/public_catalog.py — browse and filter the public catalog
• samples/core/benchmark_evaluation.py — run model vs. benchmark and retrieve results

Stratix pattern

1. Browse the Stratix Public catalog to shortlist 3-5 candidates.
2. Use client.public.models.get(query=, name=, key=, categories=, companies=, regions=, licenses=) to filter programmatically. Discover available filter values via response.categories, response.companies, etc.
3. Run a private evaluation against your dataset using client.evaluations.create(model=, benchmark=).
4. Read the per-dimension result deltas, not just headline accuracy.

Where to next

• Use case: Model evaluation
• Tutorial 1: First evaluation
• Compare models (Stratix Public)