Use Case Wizard
4 questions and we recommend the top 3 best-fit models for your task, with benchmark-based rationale. Explicit rules — no black box.
Step 1 of 4
What do you want to do?
How we recommend
Each use case has signal benchmarks with weights (1.0 primary, 0.5 secondary). The composite score is:
score = sum(weight × benchmark_score) / sum(present weights)
composite = score × (0.4 + 0.6 × coverage) × priority_factor
Coverage penalizes models with little published data. priority_factor adjusts according to your choice: price penalizes expensive models, speed favors mini/flash variants, quality is neutral.
It is rules-based, not ML. The rules live in the code and are reviewable. No editorial magic.