Open data

Everything on this website is generated from these files. Download them, build your own charts, or check our numbers. The raw per-pass result bundles live in the repository.

Downloads

  • index.json

    Project identity and the list of domains with their headline metrics.

    JSON · 2.1 KiB · benchcore.dataset/v1

  • backend.json

    Backend services: metric registry, 12 scenarios, 15 implementations, machines and 2 runs with summaries.

    JSON · 626.4 KiB · benchcore.dataset/v1

Generated 19 Nov 2026, 21:16 UTC.

License

The data is licensed under CC-BY-4.0 . You may share and adapt it for any purpose if you give appropriate credit and indicate changes. The code is licensed under MIT.

How to cite

Benchmania (2026). Community-driven, reproducible performance benchmarks. https://benchmania.cleson.dev/. Data licensed under CC-BY-4.0.

Raw result bundles

Each run is one bundle with the machine fingerprint, CPU layout, every pass's raw values and the summary. CI checks that the summary matches the per-pass data.

Browse all bundles

Data contract

The shape of the dataset files (schema benchcore.dataset/v1). Results are summaries over passes: m is the median, lo/hi the 95 % bootstrap interval of the median, n the number of passes and cv the coefficient of variation.

index.json
  schema, generated, project { name, slug, tagline, description, repository, site_url, license, data_license }
  domains[] { id, title, summary, status, service, headline[], description, runs, default_run, ... }

<domain>.json
  metrics       { <metricId>: { label, unit, better, group, description, decimals, visibility } }
  derived[]     { id, kind, metric, scenarios }
  scenarios[]   { id, title, kind, order, description, params }
  implementations[] { id, name, language, runtime, framework, variant?, category, versions, notes, source? }
  machines      { <machineId>: { label, reference?, cpu, memory_gib, kernel, os, virtualization, docker } }
  runs[]        { id, created, finished, status, methodology, provenance, machine, passes,
                  options, layout, tool, flags[],
                  implementations { <implId>: { image, check, manifest } },
                  results { <implId>: { <scenarioId>: { <metricId>: { m, lo, hi, min, max, n, cv } } } },
                  derived { <implId>: { <derivedMetricId>: number } } }
  default_run