> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hitheo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Get Benchmarks

> Retrieve aggregated platform performance benchmarks for a given time window.

## Query Parameters

<ParamField query="window" type="string" optional>
  Time window for aggregation. One of: `24h`, `7d`, `30d`. Default: `24h`.
</ParamField>

## Response

<ResponseField name="window" type="string">
  The requested time window.
</ResponseField>

<ResponseField name="latency_by_mode" type="object">
  Per-mode latency statistics: `{ [mode]: { p50, p95, p99, count } }`.
</ResponseField>

<ResponseField name="ttft_by_mode" type="object">
  Per-mode time-to-first-token statistics (streaming only): `{ [mode]: { p50, p95, count } }`.
</ResponseField>

<ResponseField name="classification" type="object">
  Intent classification latency: `{ p50, p95, count }`. Null if no auto-mode requests in the window.
</ResponseField>

<ResponseField name="counters" type="object">
  Aggregate counters: `total`, `failover_count`, `routing_accuracy`, `total_tokens`, `total_cost_cents`, `by_mode`.
</ResponseField>

## Authentication

Requires an API key or a dashboard session.

## Example

```bash curl theme={null}
curl "https://www.hitheo.ai/api/v1/benchmarks?window=24h" \
  -H "Authorization: Bearer $THEO_API_KEY"
```

### Response

```json theme={null}
{
  "window": "24h",
  "latency_by_mode": {
    "fast": { "p50": 450, "p95": 1200, "p99": 2100, "count": 1543 },
    "code": { "p50": 890, "p95": 2400, "p99": 4100, "count": 312 }
  },
  "ttft_by_mode": {
    "fast": { "p50": 180, "p95": 450, "count": 1200 }
  },
  "classification": { "p50": 12, "p95": 310, "count": 890 },
  "counters": {
    "total": 2847,
    "failover_count": 3,
    "routing_accuracy": 0.94,
    "total_tokens": 4520000,
    "total_cost_cents": 1250,
    "by_mode": { "fast": 1543, "code": 312, "image": 201 }
  },
  "bucket": "2026-04-16T05"
}
```