Open data

The standings are free to read, reuse and cite, by people and by AI agents. Plain JSON over HTTPS, no key, CORS open (Access-Control-Allow-Origin: *).

Endpoints

Data endpoints
URL	What
https://airacescoreboard.com/data/v1/standings.json	This week's standings: rank, score, components, model ranks, revenue, valuation, odds, weekly changes, plus the weights, formula and source timestamps.
https://airacescoreboard.com/data/v1/labs.json	Every sourced fact per lab: revenue (reported and claimed), valuation (closed and in talks), total funding, compute and dated receipts. Each fact has value, basis, as_of and source_url.
https://airacescoreboard.com/data/v1/labs/{id}.json	One lab: its current standing plus its sourced facts. Ids: `anthropic`, `google`, `openai`, `meta`, `alibaba`, `xai`, `moonshot`, `zai`, `deepseek`, `mistral`.
https://airacescoreboard.com/data/v1/history/index.json	List of weekly snapshots, newest first.
https://airacescoreboard.com/data/v1/history/{YYYY-MM-DD}.json	One weekly snapshot in the same standings shape.
https://airacescoreboard.com/feed.xml	RSS 2.0 feed, one item per weekly snapshot.
https://airacescoreboard.com/llms.txt	Short plain-text summary for AI agents (llmstxt.org format).
https://airacescoreboard.com/llms-full.txt	Everything in one plain-text file: methodology, current standings and every sourced fact.

Example

curl https://airacescoreboard.com/data/v1/standings.json

{
  "schema_version": 1,
  "date": "2026-10-17",
  "standings": [
    {
      "rank": 1,
      "id": "anthropic",
      "name": "Anthropic",
      "score": 94.4,
      "coverage": 5,
      "components": {
        "arena": 84,
        "aa": 100,
        "revenue": 100,
        "valuation": 100,
        "odds": 100
      },
      "prev_rank": 2,
      "rank_change": 1,
      "score_change": 0.5,
      "arena": {
        "rank": 3,
        "score": 1508,
        "model": "Claude Opus 5.5 (High)"
      },
      "artificial_analysis": {
        "index": 57.6223698102963,
        "position": 1,
        "model": "Claude Opus 5.5 (Max, Default Fallback)"
      },
      "revenue_usd": 65000000000,
      "valuation_usd": 965000000000,
      "polymarket_odds": 0.475,
      "odds_change_pp": 6,
      "url": "https://airacescoreboard.com/lab/anthropic/",
      "data_url": "https://airacescoreboard.com/data/v1/labs/anthropic.json"
    }
  ]
}

Fields

Standings fields
Field	Meaning
rank	Position this week (1 = leader).
score	Composite score, 0-100. null if no component has data.
coverage	How many of the 5 components had data (e.g. 3 = 3/5).
components	Each component scaled 0-100: arena, aa, revenue, valuation, odds. null = no reliable public data, left out of the score.
rank_change / score_change	Change vs the snapshot about 7 days earlier (positive = up). null for a new entry.
arena	Best Arena text leaderboard row for the lab: rank, score, model.
artificial_analysis	Best Artificial Analysis Intelligence Index for the lab: index, position, model.
revenue_usd / valuation_usd	The scored figures, in US dollars. Basis, date and source are in labs.json.
polymarket_odds	Yes price (0-1) in the scored Polymarket market.
odds_change_pp	Weekly odds change, percentage points.

How each component is calculated is on the methodology page, and in the weights and formula fields of every standings file.

Update cadence

Weekly, on Mondays (Europe/Bucharest). Polymarket and Artificial Analysis are fetched on each run; Arena is a manual snapshot; revenue, valuation and funding are updated by hand when a sourced figure changes. The date field is the run day; generated_at and sources.*.fetched_at are exact timestamps. Responses are cached for up to 5 minutes.

Versioning

The /data/v1/ shape is stable. It may gain new fields, but existing fields keep their names and meaning. A breaking change gets a new path (/data/v2/) and v1 stays up. Every file has schema_version.

License

The standings data is licensed under CC BY 4.0. You may copy, share and adapt it, also commercially, if you credit AI Race Scoreboard with a link to https://airacescoreboard.com. Every JSON file repeats this in its license field.

How to cite

Link to the page you used and give the week, for example:

AI Race Scoreboard, weekly AI lab standings, week of 2026-10-17. https://airacescoreboard.com/

Each lab file has its own citation line. The underlying figures come from Arena, Artificial Analysis, Polymarket and the outlets linked in source_url: credit them too.