Get Statistics Performance
Troubleshooting
Encountering errors? See the API Error Codes Reference for a complete list of error types and troubleshooting steps.
GET https://openqiao.com/api/v1/management/statistics/performanceRank the top 10 models by a performance metric over a date range. Supports throughput, time-to-first-token latency, and cache hit rate. Each model entry breaks the metric down by provider so you can compare providers for the same model.
Use this endpoint to build performance leaderboards, pick the fastest provider for a model, or monitor cache efficiency.
Data freshness
Statistics data is aggregated on a daily schedule. The most recent available data is from yesterday (T-1). Today's usage will appear in tomorrow's aggregation.
Authentication
Authorization Header *
Authorization: Bearer <OPENQIAO_MANAGEMENT_API_KEY>- Name:
Authorization - Format:
Bearer <API_KEY> - Description: A Management API Key created in the 开桥 Console
Management API Key required
This endpoint only accepts Management API Keys. Standard API Keys are not supported.
Rate Limiting
Each endpoint has its own independent rate limit counter. The maximum number of requests per minute is configured at the platform level. Exceeding the limit returns a 422 error.
Parameters
metric string *
Which performance metric to rank by.
throughput— median output throughput (tokens/sec), sorted descendingttft— median time-to-first-token latency (ms), sorted ascendingcache_hit— cache hit rate, sorted descending
starting_at string *
Start date (inclusive), in YYYY-MM-DD format. Clamped to 2025-09-29 (platform data floor).
ending_at string *
End date (inclusive), in YYYY-MM-DD format. Must not be later than today.
Constraints
- Date range must not exceed 30 days. Exceeding returns
400. starting_atmust not be later thanending_at.- For
throughput/ttft, days with no valid samples (median0, 例如 embedding/TTS/video models) are excluded, so models without latency/throughput data are not pushed to the top.
Returns
data.metric string
Echo of the requested metric ("throughput", "ttft", or "cache_hit").
data.starting_at string
Effective start date (YYYY-MM-DD).
data.ending_at string
Effective end date (YYYY-MM-DD).
data.entries array
Top 10 models. The entry shape depends on metric.
For throughput / ttft:
rankinteger— 1-based rankmodelstring— Model sluglabelstring— Model display nameicon_urlstring— Model (or author fallback) icon URLprovidersarray— Per-provider breakdown (sorted best-first):providerstring— Provider slugprovider_labelstring— Provider display nameprovider_icon_urlstring— Provider icon URLvaluenumber— Metric value (throughput tokens/sec, or latency ms)value_labelstring— Pre-formatted label (例如,"152.34"or"420ms")
For cache_hit:
rankinteger— 1-based rankmodelstring— Model sluglabelstring— Model display nameicon_urlstring— Model (or author fallback) icon URLtop_providerobject— The (provider, api) pair with the highest hit rate:provider_labelstring— Provider display nameapistring— API surface (例如,anthropic,openai)valuenumber— Hit rate (0–1)value_labelstring— Pre-formatted percentage (例如,"86.42%")
providersobject— Map keyed by provider slug:provider_labelstring— Provider display nameprovider_icon_urlstring— Provider icon URLapisarray— Per-api hit rate (sorted descending):apistring— API surfacevaluenumber— Hit rate (0–1)value_labelstring— Pre-formatted percentagecache_tokensnumber— Cache-read tokens (input_cache_read)prompt_tokensnumber— Hit-rate denominator tokens (read + prompt + all cache-write tiers)
curl -G https://openqiao.com/api/v1/management/statistics/performance \
-H "Authorization: Bearer $OPENQIAO_MANAGEMENT_API_KEY" \
-d metric=throughput \
-d starting_at=2026-06-14 \
-d ending_at=2026-06-21{
"success": true,
"data": {
"metric": "throughput",
"starting_at": "2026-06-14",
"ending_at": "2026-06-21",
"entries": [
{
"rank": 1,
"model": "qwen/qwen3-coder",
"label": "Qwen3-Coder",
"icon_url": "https://cdn.marmot-cloud.com/storage/openqiao/2026/04/01/qeMamJm/Property-1Qwen.svg",
"providers": [
{
"provider": "alibaba",
"provider_label": "Alibaba",
"provider_icon_url": "https://cdn.marmot-cloud.com/storage/openqiao/2026/04/01/qeMamJm/Property-1Qwen.svg",
"value": 152.34,
"value_label": "152.34"
}
]
}
]
}
}{
"success": true,
"data": {
"metric": "cache_hit",
"starting_at": "2026-06-14",
"ending_at": "2026-06-21",
"entries": [
{
"rank": 1,
"model": "anthropic/claude-sonnet-4",
"label": "Claude Sonnet 4",
"icon_url": "https://cdn.marmot-cloud.com/storage/openqiao/2025/10/15/xxx/Property-1Anthropic.svg",
"top_provider": {
"provider_label": "Anthropic",
"api": "anthropic",
"value": 0.8642,
"value_label": "86.42%"
},
"providers": {
"anthropic": {
"provider_label": "Anthropic",
"provider_icon_url": "https://cdn.marmot-cloud.com/storage/openqiao/2025/10/15/xxx/Property-1Anthropic.svg",
"apis": [
{
"api": "anthropic",
"value": 0.8642,
"value_label": "86.42%",
"cache_tokens": 128450000,
"prompt_tokens": 148630000
}
]
}
}
}
]
}
}