# 模型坐标：完整机器说明

## Identity
模型坐标是中文 AI Benchmark 信息聚合站。本站主要收集并结构化展示外部公开评测，不修改来源分数，也不把不同单位合成为本站总分。少数中文专题会对既有输出运行单独披露的 AI 复评，所得分数只在该固定样本专题内使用。

## Provenance
- Primary source: https://artificialanalysis.ai/
- Homepage snapshot: 2026-07-21
- Model dataset version: 472f88c74bc2
- Model dataset snapshot: 2026-07-22
- Every editorial page contains a visible source link and snapshot date.
- Chinese evaluation source: https://github.com/k2009/Kline-star/blob/main/docs/ai-model-evals/tencentmaas-yearly-report/compare-full-model-set-2026.md
- Chinese evaluation snapshot: 2026-07-22; it preserves a historical fixed-sample evaluation and is not a current universal model ranking.
- Fortune-telling content score: site-generated blind AI review; the judge model, rubric version, per-run scores, disagreement handling, and source output URLs are preserved in the Chinese evaluation dataset.

## Canonical entry points
- Homepage: https://models.latent.vip/
- Model leaderboard: https://models.latent.vip/models/
- Full model directory (580 routes): https://models.latent.vip/models/directory/
- Country analysis: https://models.latent.vip/countries/
- Coding agents: https://models.latent.vip/coding-agents/
- Media leaderboards: https://models.latent.vip/media/
- Capability indices: https://models.latent.vip/capabilities/
- Chinese evaluations: https://models.latent.vip/chinese-evaluations/
- Fortune-telling yearly-detail evaluation: https://models.latent.vip/chinese-evaluations/fortune-telling/
- Evaluation views: https://models.latent.vip/evaluations/
- Cost and performance: https://models.latent.vip/economics/
- API providers: https://models.latent.vip/providers/
- Hardware: https://models.latent.vip/hardware/
- Methodology: https://models.latent.vip/methodology/
- Update archive: https://models.latent.vip/updates/

## Machine-readable endpoints
- Data catalog: https://models.latent.vip/data/
- Homepage snapshot: https://models.latent.vip/data/homepage.json
- Model metrics: https://models.latent.vip/data/models.json
- Sync change manifest: https://models.latent.vip/data/changes.json
- Chinese evaluation dataset: https://models.latent.vip/data/chinese-evaluations.json
- RSS updates: https://models.latent.vip/feed.xml
- Sitemap: https://models.latent.vip/sitemap.xml

## Interpretation rules
1. Preserve the exact model variant and reasoning-effort label from the source.
2. Do not compare metrics with different units as if they were the same score.
3. Estimated or provisional values remain marked as such.
4. Treat this site as a discovery and explanation layer; cite the linked primary source for final claims.
5. Chinese evaluation scores remain scoped to their fixed samples and disclosed judge protocol; do not reinterpret them as a cross-benchmark total score.