---
title: "Get a benchmark by slug"
method: GET
path: "/api/benchmarks/{slug}"
tags: ["benchmarks"]
---

# Get a benchmark by slug

`GET /api/benchmarks/{slug}`

## Request

- Base URL: `https://openbenchmarks.com`
- URL: `https://openbenchmarks.com/api/benchmarks/{slug}`
- Auth: none declared

Returns full per-provider results for a benchmark. Safe to call without authentication.

## Path parameters

- `slug` 'lookalikes' | 'voice-agent-latency' | 'text-to-speech-benchmark-by-coval' | 'speech-to-text-benchmark-by-coval', required

## Response `200`

Full benchmark data.

- Benchmark
  - `slug` string, required
  - `name` string, required
  - `description` string
  - `page_url` string, uri
  - `last_updated` string, date-time
  - `license` string, uri
  - `metrics` Metric[]
    - `key` string, required
    - `label` string, required
    - `unit` 'percent' | 'usd' | 'milliseconds' | 'count', required
    - `direction` 'higher_is_better' | 'lower_is_better' | 'neutral', required
    - `definition` string
  - `categories` string[]
  - `active_dataset` string, nullable
  - `results` union[], required: Per-provider results. Lookalike rows report Precision@10/50/100 per vendor; voice (by-Coval) rows report per-model latency percentiles (ms) and Word Error Rate (percent).
    - union
      - LookalikeProviderResult: Per-vendor row on the lookalike benchmark. Precision is fixed-cutoff (10 / 50 / 100); Precision@100 is the headline.
        - `rank` integer, required
        - `provider` string, required
        - `provider_slug` string, required
        - `avg_precision_at_10` number, nullable: Mean Precision@10 across judged seeds (share of the top 10 returned the LLM judge scored relevant).
        - `avg_precision_at_100` number, nullable: HEADLINE metric. Mean Precision@100 across judged seeds.
        - `total_relevant` integer, nullable: Total relevant lookalikes the judge scored across the cohort. Reach / tiebreak metric.
        - `seeds_judged` integer, nullable: Seeds where the vendor returned enough results and the judge scored every one.
        - `total_seeds` integer, nullable
        - `cost_per_relevant_usd` number, nullable: Estimated request spend divided by total relevant. Null when public pricing is unavailable.
        - `avg_latency_ms` number, nullable
      - VoiceModelResult: Per-model row on a voice benchmark (by Coval). Latency values are milliseconds; WER is percent; lower is better on both. Voice quality/naturalness and price are not measured.
        - `rank` integer, required
        - `provider` string, required
        - `model` string, required
        - `latency_metric` string: Which latency is reported: TTFA (TTS, time to first audio) or TTFS (STT, time to final segment).
        - `latency` VoiceMetricStats: Distribution stats for one metric of one model, aggregated over the rolling window.
          - `unit` 'ms' | 'percent' | 'ratio'
          - `avg` number, nullable
          - `p50` number, nullable
          - `p90` number, nullable
          - `p95` number, nullable
          - `p99` number, nullable
          - `min` number, nullable
          - `max` number, nullable
          - `samples` integer, nullable
        - `wer` VoiceMetricStats: Distribution stats for one metric of one model, aggregated over the rolling window.
          - `unit` 'ms' | 'percent' | 'ratio'
          - `avg` number, nullable
          - `p50` number, nullable
          - `p90` number, nullable
          - `p95` number, nullable
          - `p99` number, nullable
          - `min` number, nullable
          - `max` number, nullable
          - `samples` integer, nullable
  - `methodology_url` string, uri

## Other responses

- `404`: Benchmark slug not recognized.

---

[API](https://skmtc.dev/openbenchmarks/apis/openbenchmarks-public-api.md) - [All operations](https://skmtc.dev/openbenchmarks/apis/openbenchmarks-public-api/llms.txt) - [OpenAPI document](https://skmtc.dev/openbenchmarks/apis/openbenchmarks-public-api/revisions/f6106dbd118d?raw)