Qwen · Search Reranking · Open Weight

Decision checklist

Check model-specific API compatibility; evaluate quality on your own data; confirm commercial terms; enforce usage caps; secure credentials; monitor cost and latency.

• candidate ordering schema
• nDCG and precision at k
• retrieval/rerank separation
• qwen3.7-text-embedding — Embeddings and retrieval; vector schema
• qwen3.7-text-embedding-flash — Embeddings and retrieval; vector schema

Practical implementation

Measure nDCG and precision at k; rerank only authorized candidate passages.

Measure task accuracy, p50/p95 latency, token use, error rate, tool-call reliability, and total cost on a fixed test set. Check model-specific API compatibility; evaluate quality on your own data; confirm commercial terms; enforce usage caps; secure credentials; monitor cost and latency.

Compare input and output token volume, context behavior, caching, retries, latency, and infrastructure overhead before quoting. No unverified price or savings claims are published.

Sample evaluation prompt

Task: search-reranking. Model candidates: qwen3.7-text-rerank, qwen3.7-text-embedding, qwen3.7-text-embedding-flash. Write a response strictly based on the supplied facts. If evidence is missing, answer 'insufficient evidence'. Return JSON keys answer, evidence, confidence_note.

This code is illustrative, not a live API endpoint. Your approved base URL and credentials are supplied only after provisioning.

Cost planning

Compare input and output token volume, context behavior, caching, retries, latency, and infrastructure overhead before quoting. No unverified price or savings claims are published.

Measure task accuracy, p50/p95 latency, token use, error rate, tool-call reliability, and total cost on a fixed test set.

Related Qwen resources