Methodology & Trust

Recommendation Scoring (Phase A v1)

Model Finder ranks candidates by a weighted fit score and returns both reasons and tradeoffs. The current breakdown has five components:

• Task fit: keyword and capability matching for your selected task
• Cost fit: budget alignment based on current input and output pricing
• Context fit: minimum context requirement check
• Priority fit: cheapest, fastest, quality-first, or balanced preference
• Intent fit: optional free-text signals (JSON output, speed, multilingual, reliability)

Data Freshness

Pricing data is refreshed from provider and aggregator sources on recurring schedules. Recommendation results include a visible pricing refresh timestamp.

If your workload is highly sensitive to price or latency, validate with a direct provider quote and a short canary benchmark before production rollout.

Confidence & Limitations

• Recommendations are guidance, not guarantees of business outcomes.
• Latency and reliability can vary by region, load, and provider routing.
• Some model metadata may be incomplete or changed by providers without notice.
• Use the returned tradeoffs section to understand where additional validation is required.