Get benchmark
GET/v1/benchmark
Recorded runs of a fixed question set put to this product's Ask and to Claude Opus 5.5 with web search: per question, each side's verdict against the golden answers, latency, and cost, and the product's status. Only runs measured from a cloud region on the launch hosting are served, and no comparison text. Public.
- scope
- none
- auth
- none
- operationId
- GetBenchmark
Response
200 · BenchmarkResponse. The published runs, newest first.
| Field | Type |
|---|---|
| runs | BenchmarkRun[] |
Problems
RFC 9457 application/problem+json. An abstention is not one of these: it is a 200.
| HTTP | Code | When |
|---|---|---|
| 503 | STORE_UNAVAILABLE, LLM_UNAVAILABLE, LOG_UNAVAILABLE | A dependency is down or a budget tripped |
| 504 | TIMEOUT | A deadline was exceeded |