AI attention
A versioned public prompt corpus is sampled across supported answer engines. Recommendation frequency, prominence, citations, category breadth, and provider breadth contribute to visibility.
Measure the signal.
Show the source.
Every public ranking is a versioned snapshot—not an opaque live score. This page describes the principles the production data pipeline is designed to enforce.
What makes public rankings trustworthy
Compute markets
Verified rental listings are normalized to price per physical GPU hour. Accelerator model, memory, topology, contract type, geography, and availability are preserved.
Public equity prices
Licensed end-of-day closes are attached only to reviewed exchange-traded securities. The displayed move compares two available closes; it is not an intraday quote.
Private-company valuations
A private valuation appears only when the company or transaction participants disclose a financing valuation. It is labeled by transaction date and is never presented as a stock price.
Prediction signals
Prediction-market prices remain raw market signals unless a versioned, reviewable transformation defines an implied valuation. They are never mixed with disclosed transaction valuations.
Company financials
Public filing facts are mapped into a versioned financial taxonomy. Debt, cash, leases, EBITDA, and capital expenditure retain source period and filing provenance.
Model economics
Published API pricing and measured influence are separate metrics. Model influence does not represent intelligence, safety, or benchmark quality.
Confidence
Every score includes observation coverage, source freshness, provider success, and entity-resolution confidence. Low-confidence values are withheld from ranking.
Corrections
Entity merges, source corrections, and methodology changes are logged. Historical snapshots remain tied to the methodology version used at publication.
LIVE BOARD LEDGER
Published criteria, not mystery scores
Only boards with a publishable verified snapshot appear here. Every row links to its current evidence ledger and frozen methodology.
| Board | Primary criterion | Ranked | Confidence | Freshness | Published measurement rule | Evidence |
|---|---|---|---|---|---|---|
| Open Source AIopen-source-ai | GitHub stars | 14 | 100% | current | Repository stargazers_count reported by GitHub. Daily cumulative snapshots are the input for trailing star-growth calculations. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 99% mean confidence. At least 14 entities must clear the quality gate. | Inspect leader → |
| AI Model & API Brand Rankai-model-api-brand-rank | Recommendation-prompt presence | 17 | 100% | current | Percent of successful benchmark answers to recommendation-intent prompts that mention or cite the brand. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 17 entities must clear the quality gate. | Inspect leader → |
| AI Research Activityai-research | OpenAlex AI research activity score | 10 | 100% | current | A 0–100 cohort-relative composite of indexed work count, current citations to those works, and citations per work. It measures OpenAlex-indexed activity, not total research output or quality. Rows are ordered by the published openalex-ai-research-activity formula v1; ties retain deterministic entity ordering. Refresh: daily. Publication requires at least 99% mean confidence. At least 10 entities must clear the quality gate. | Inspect leader → |
| Model Adoption Signalsmodel-adoption | Hugging Face downloads (30 days) | 12 | 100% | current | The Hub model-info downloads field: repository downloads over the trailing 30 days. It is not unique users or global market share. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 12 entities must clear the quality gate. | Inspect leader → |
| Countries in AI Researchcountries-in-ai-research | OpenAlex-indexed AI works by authorship country | 216 | 100% | current | Distinct OpenAlex works in the trailing 365-day AI scope with at least one authorship assigned to the country. A multinational work counts once in each represented country. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 99% mean confidence. At least 5 entities must clear the quality gate. | Inspect leader → |
| AI Developer Packagesai-developer-packages | npm downloads (trailing 30 days) | 10 | 100% | current | npm downloads (trailing 30 days) for one reviewed public npm package. Counts are not unique users or global adoption. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 10 entities must clear the quality gate. | Inspect leader → |
| Official AI Infrastructure Investment Disclosuresofficial-ai-infrastructure-investment | Disclosed infrastructure investment reference amount | 8 | 100% | current | The headline project investment reference amount in a reviewed first-party announcement. It is not actual spend; the announcement qualifier is mandatory metadata. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 8 entities must clear the quality gate. | Inspect leader → |
| States in Reviewed AI Infrastructure Announcementsstates-in-reviewed-ai-infrastructure-announcements | Reviewed announced AI infrastructure projects | 6 | 100% | current | Count of distinct reviewed project identities in official company announcements mapped to the state. Rows are ordered by the board's published primary metric. Refresh: daily_after_source_verification. Publication requires at least 100% mean confidence. At least 6 entities must clear the quality gate. | Inspect leader → |
| AI Agent Framework Activityai-agent-framework-activity | npm downloads (trailing 30 days) | 8 | 100% | current | npm downloads (trailing 30 days) for one reviewed public npm package. Counts are not unique users or global adoption. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 8 entities must clear the quality gate. | Inspect leader → |
| AI Company Financialscompany-financials | Company financial scale score | 11 | 100% | current | Cohort-relative financial scale derived from reported revenue, assets, capital expenditure, cash, and disclosed AI/cloud revenue. Rows are ordered by the published financial-log-minmax-weighted-scale formula v1; ties retain deterministic entity ordering. Evidence: sec-companyfacts, company-ir. Refresh: daily. Publication requires at least 75% mean confidence. At least 3 entities must clear the quality gate. | Inspect leader → |
| Reported Capital Expenditurereported-capex | Reported capital expenditure scale | 10 | 100% | current | Cohort-relative log-normalized score backed by the latest eligible annual reported capital-expenditure observation. Rows are ordered by the published reported-capex-log-minmax formula v1; ties retain deterministic entity ordering. Evidence: sec-companyfacts. Refresh: daily. Publication requires at least 90% mean confidence. At least 3 entities must clear the quality gate. | Inspect leader → |
| Forward Capital Expenditureforward-capex | Forward capital expenditure scale | 3 | 100% | current | Cohort-relative log-normalized score backed only by current first-party capital-expenditure guidance. Rows are ordered by the published forward-capex-log-minmax formula v1; ties retain deterministic entity ordering. Evidence: company-ir. Refresh: daily. Publication requires at least 80% mean confidence. At least 3 entities must clear the quality gate. | Inspect leader → |
| Debt Exposuredebt-exposure | Debt exposure score | 7 | 100% | current | Cohort-relative balance-sheet exposure derived from reported current and noncurrent long-term debt divided by total assets. A higher score means greater exposure, not better credit quality. Rows are ordered by the published debt-to-assets-winsorized-minmax formula v1; ties retain deterministic entity ordering. Evidence: sec-companyfacts. Refresh: daily. Publication requires at least 90% mean confidence. At least 3 entities must clear the quality gate. | Inspect leader → |
| AI Stack Controlai-stack-control | AI stack control score | 7 | 100% | current | Evidence-backed coverage of reviewed silicon, model, cloud, data-center, distribution, and open-weight capabilities. Rows are ordered by the published ai-stack-known-capability-share formula v1; ties retain deterministic entity ordering. Evidence: company-ir. Refresh: daily. Publication requires at least 75% mean confidence. At least 3 entities must clear the quality gate. Missing evidence is treated as unknown, never as a negative. | Inspect leader → |
| Azure GPU VM Retail Pricesazure-gpu-vm-prices | Normalized retail price per GPU | 6 | 100% | current | VM retail price divided by the documented physical GPU count. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 6 entities must clear the quality gate. | Inspect leader → |
| AI Accelerator Memory Specificationsai-accelerator-memory-specs | Accelerator memory capacity | 5 | 100% | current | Vendor-reported accelerator memory capacity. Rows are ordered by the board's published primary metric. Refresh: event_driven_with_daily_review_check. Publication requires at least 100% mean confidence. At least 5 entities must clear the quality gate. | Inspect leader → |
| AI Dataset Activityai-datasets | Hugging Face dataset downloads (30 days) | 12 | 100% | current | Downloads reported by the Hugging Face dataset-info API for the repository. This is platform activity, not unique users or quality. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 12 entities must clear the quality gate. | Inspect leader → |
| Federal AI Contract Obligationsgovernment-ai-contract-obligations | Federal AI-described transaction obligations | 108 | 100% | current | Federal action obligations on contract transactions whose descriptions and industry/product codes pass the published classification contract. This is not potential award value. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 3 entities must clear the quality gate. | Inspect leader → |
| Federal AI Obligations by Awarding Agencyfederal-ai-obligations-by-agency | Federal AI-described transaction obligations | 14 | 100% | current | Federal action obligations on contract transactions whose descriptions and industry/product codes pass the published classification contract. This is not potential award value. Rows are ordered by the board's published primary metric. Refresh: daily. Publication requires at least 100% mean confidence. At least 3 entities must clear the quality gate. | Inspect leader → |
Commercial relationships do not determine corpus membership, scoring, or rank. Rank.ai customers receive analysis and recommendations—not favorable public scores.
Customer prompts remain private
Public rankings use a separately governed Rank.ai benchmark corpus. Private customer prompts and competitive sets are never published or used to identify public ranking entities without explicit consent.
Context, not investment advice
Stock, debt, and capital-expenditure data is presented as factual market context. Rank.ai scores do not predict securities performance and should not be treated as investment recommendations.