China AI Hub Data
This is the data layer for China AI Hub — the English-language structured information layer for China's AI ecosystem.
While the main site hosts editorial content, comparisons and original research, this data hub is dedicated to the structured reference data behind it: what is known, verified, and sourceable about Chinese AI models, companies, agents, APIs, pricing and benchmarks.
What this hub covers
| Section | Scope |
|---|---|
| Models | Chinese foundation and coding models: capabilities, context windows, training knowledge, release dates |
| Companies | The organizations building China's AI stack: ownership, funding, products |
| Agents | Consumer and developer agent products: features, platforms, defaults |
| APIs | API platforms and endpoints: access, modalities, token limits |
| Pricing | Time-stamped pricing snapshots with effective dates, plus documented price_history |
| Benchmarks | Benchmark methodologies and vendor-reported results |
Data schemas
- Entity Template — unified field structure for every collection
- Evidence Layer — unified
sourcesschema (evidence_id,source_type,published,verified,confidence,conflict) - History Data — structured historical dimensions (
model_versions,license_changes,api_changes,company_milestones,benchmark_changes)
Historical dimensions
Beyond current-value fields, each collection carries sourced history where documented: models expose a release_history timeline; pricing exposes price_history change events; companies expose a timeline of founding, release and pricing milestones. Every history entry includes a source and verification_date. benchmark_history is intentionally not built — no sourced before/after benchmark values exist for any model+benchmark pair.
Relationship to the main site
- Main site (sinoaihub.com) = editorial authority: what it means, how to choose, original research
- Data hub (data.sinoaihub.com) = reference layer: what is known, with sources
Every entity here links back to its canonical page on the main site. All data follows the same source policy as the main site: no fabricated data, official sources first, FACT / VENDOR CLAIM / ANALYSIS classification, and unknown where information is not publicly disclosed.
Status
The data hub is fully migrated from the main site's structured collections and is maintained as the canonical structured-fact + evidence + history layer. Every entity page carries the complete eight-element block (definition, key facts, relationships, evidence, source history, update date, related entities, and a canonical main-site link), with sources and last_verified dates on every record.
- Models: 21 tracked
- Companies: 12 tracked
- Agents: 18 tracked
- APIs: 6 tracked
- Pricing: 6 providers tracked
- Benchmarks: 18 tracked
Completeness self-check
Entity counts are verified across four surfaces by scripts/check-entity-counts.py in the main-site repo: main-site content files, data-hub records, the main-site sitemap, and the data hub's declared entities (nav + index tables) must all agree. Any drift fails the check. See the main repo's docs/data-integrity-check.md for the current comparison table and the mechanics of the check.
- Eight-element coverage — all 81 entity pages (21 models + 12 companies + 18 agents + 6 APIs + 6 pricing + 18 benchmarks) carry Definition/Description, Key facts, Sources, and a canonical main-site link; source-history notes are present where applicable.
- Canonical links — every record links to its main-site page; dangling canonicals are fixed as part of migration.
- Cross-surface parity — models, companies, agents, APIs, pricing and benchmarks each show 0 discrepancy between the four surfaces.