Open Data · CC BY 4.0
Chinese AI Agent open dataset
OKCodex publishes an AI Agent catalog, source-based reviews, comparisons, demand signals, and human-verified benchmark runs. Records label their evidence basis, version, sources, date, and automatic/manual publication mode. Some source material is in Chinese.
Catalog scores (0–100) are information assessments. Published reviews use a separate five-dimension scale (1–5). Automatic reviews analyze public sources and are not hands-on tests. Benchmark records describe specific human-verified runs. Do not convert these measures into one ranking.
60 catalog agents · 12 with qualifying review records
Follow a record to its evidence
Keep source-based reviews separate from measured runs. Check the task criteria, model version, outcome and date before reusing a benchmark claim.
Download the original data
Interpret the evidence fields
codex_score: 0–100 catalog assessment, not measured performance.review.overalland five dimensions: 1–5 review scores with supporting reasons.publication_mode: automatic source analysis or manual publication; inspect actual evidence before claiming a hands-on test.source_url / tested_version / reviewed_at: source, product/version recorded in the source and review date; not an executed model version.canonical / fit_for / not_fit_for: source page and suitability boundaries.- Missing measurements remain missing; do not infer zero cost, zero failures, or an untested result.
License, attribution and freshness
Data is available under CC BY 4.0. Attribute OKCodex (www.okcodex.com), link the specific canonical page and source, and retain version, date and evidence type.
Use record-level dates for freshness. Prices and availability must be checked against vendor sources before purchasing. This directory is curated and does not claim complete market coverage.
GitHub dataset mirror, schema and license