For AI agents

Rank and score of every model on about 30 public AI leaderboards, captured daily, plus a composite that names the leader of leaders.

Updated 13h agoCaptured every day at 04:30 UTC

Point your agent here

Summary for language modelshttps://arena.fru.dev/llms.txt
Full data as plain texthttps://arena.fru.dev/llms-full.txt
OpenAPI 3.1 spec (load as tools)https://arena.fru.dev/openapi.json

Call the API

MethodPathParamsReturns

GETGET /api/leaderslimit?, offset?models in rank order with podium score (0 to 100), #1 finishes, top-10 places and median rank
GETGET /api/boardscategory?, limit?, offset?boards with category, metric, status (fresh, stale, retired), publish date, entry count and the #1 model and score
GETGET /api/boards/{id}id, limit?the board and up to 100 entries with rank, model, name as published, score and verified flag
GETGET /api/models/{id}idthe model (name, vendor, open weights, release date), its composite standing and one row per board
GETGET /api/searchqup to 20 ranked results with title, link and one line

GET/api/leaders

The composite ranking: every model that earns points on the fresh language-model boards

limitItems per page, 1 to 200
offsetItems to skip, for the next page

Shell

curl -s "https://arena.fru.dev/api/leaders?limit=3"

Sample response

JSON

{
  "capturedAt": "2026-09-23T13:51:56.857Z",
  "boardsCounted": [
    "lmarena-text",
    "scale-multichallenge"
  ],
  "method": "https://arena.fru.dev/method",
  "items": [
    {
      "rank": 1,
      "model": "claude-fable-5-1",
      "name": "Claude Fable 5.1",
      "vendor": "anthropic",
      "vendorName": "Anthropic",
      "openWeights": false,
      "podiumScore": 58,
      "points": 87,
      "wins": 1,
      "top10": 11,
      "listed": 11,
      "medianRank": 2,
      "url": "https://arena.fru.dev/models/claude-fable-5-1"
    },
    {
      "rank": 2,
      "model": "gpt-6-astra",
      "name": "GPT-6 Astra",
      "vendor": "openai",
      "vendorName": "OpenAI",
      "openWeights": false,
      "podiumScore": 48.7,
      "points": 73,
      "wins": 5,
      "top10": 8,
      "listed": 11,
      "medianRank": 3,
      "url": "https://arena.fru.dev/models/gpt-6-astra"
    }
  ],
  "total": 60,
  "limit": 3,
  "offset": 0,
  "next": "/api/leaders?limit=3&offset=3"
}

GET/api/boards

Every leaderboard tracked, with its current #1

categoryOnly boards in this category
limitItems per page, 1 to 200
offsetItems to skip, for the next page

Shell

curl -s "https://arena.fru.dev/api/boards?category=coding&limit=2"

Sample response

JSON

{
  "capturedAt": "2026-09-23T13:51:56.857Z",
  "items": [
    {
      "id": "lmarena-webdev",
      "name": "LMArena WebDev Arena",
      "org": "LMArena",
      "category": "coding",
      "metric": "Arena rating",
      "unit": "elo",
      "higherIsBetter": true,
      "status": "fresh",
      "publishedOn": "2026-09-22",
      "capturedAt": "2026-09-23 13:51:51",
      "entries": 100,
      "countsTowardComposite": true,
      "leader": {
        "model": "gpt-6-astra",
        "name": "GPT-6 Astra",
        "publishedAs": "gpt-6-astra-max",
        "score": 1793.1844725539563,
        "verified": true
      },
      "source": "https://lmarena.ai/leaderboard/webdev"
    },
    {
      "id": "scale-swe-pro",
      "name": "SWE-Bench Pro",
      "org": "Scale AI",
      "category": "coding",
      "metric": "Resolved",
      "unit": "%",
      "higherIsBetter": true,
      "status": "fresh",
      "publishedOn": "2026-09-23",
      "capturedAt": "2026-09-23 13:51:53",
      "entries": 11,
      "countsTowardComposite": true,
      "leader": {
        "model": "claude-opus-5",
        "name": "Claude Opus 5",
        "publishedAs": "Opus 5 (Claude Code) xhigh",
        "score": 98,
        "verified": true
      },
      "source": "https://scale.com/leaderboard/swe_bench_pro_public_v2"
    }
  ],
  "total": 4,
  "limit": 2,
  "offset": 0,
  "next": "/api/boards?category=coding&limit=2&offset=2"
}

GET/api/boards/{id}

One board: the full ranking as published, best variant per model

idBoard id from /api/boards, e.g. lmarena-text
limitEntries, 1 to 100

Shell

curl -s "https://arena.fru.dev/api/boards/lmarena-text?limit=3"

Sample response

JSON

{
  "board": {
    "id": "lmarena-text",
    "name": "LMArena Text Arena",
    "org": "LMArena",
    "category": "chat",
    "measures": "Blind pairwise votes from millions of people chatting with two anonymous models. Style control on.",
    "metric": "Arena rating",
    "unit": "elo",
    "higherIsBetter": true,
    "status": "fresh",
    "publishedOn": "2026-09-13",
    "capturedAt": "2026-09-23 13:51:50",
    "entries": 100,
    "source": "https://lmarena.ai/leaderboard/text",
    "url": "https://arena.fru.dev/boards/lmarena-text"
  },
  "ranking": [
    {
      "rank": 1,
      "model": "claude-fable-5",
      "name": "Claude Fable 5",
      "vendor": "anthropic",
      "publishedAs": "claude-fable-5-high",
      "score": 1505.682718082738,
      "verified": true,
      "capturedOn": "2026-09-23"
    },
    {
      "rank": 2,
      "model": "claude-opus-4-6",
      "name": "Claude Opus 4.6",
      "vendor": "anthropic",
      "publishedAs": "claude-opus-4-6-high",
      "score": 1504.559637415755,
      "verified": true,
      "capturedOn": "2026-09-23"
    }
  ]
}

GET/api/models/{id}

One model: its rank and score on every board that lists it

idModel id, from /api/leaders or /api/search

Shell

curl -s "https://arena.fru.dev/api/models/gpt-6-astra"

Sample response

JSON

{
  "model": {
    "id": "gpt-6-astra",
    "name": "GPT-6 Astra",
    "vendor": "openai",
    "vendorName": "OpenAI",
    "openWeights": false,
    "releaseDate": "2026-09-02",
    "url": "https://arena.fru.dev/models/gpt-6-astra"
  },
  "standing": {
    "rank": 2,
    "podiumScore": 48.7,
    "wins": 5,
    "top10": 8,
    "listed": 11,
    "medianRank": 3
  },
  "boards": [
    {
      "board": "lmarena-webdev",
      "name": "LMArena WebDev Arena",
      "category": "coding",
      "rank": 1,
      "of": 100,
      "score": 1793.1844725539563,
      "unit": "elo",
      "publishedAs": "gpt-6-astra-max",
      "verified": true,
      "status": "fresh",
      "capturedOn": "2026-09-23"
    },
    {
      "board": "scale-hle",
      "name": "Humanity's Last Exam",
      "category": "reasoning",
      "rank": 1,
      "of": 42,
      "score": 54.8,
      "unit": "%",
      "publishedAs": "GPT 6 Astra",
      "verified": true,
      "status": "fresh",
      "capturedOn": "2026-09-23"
    }
  ]
}

GET/api/search

Search models, boards, labs and pages

qA model, board or lab name

Shell

curl -s "https://arena.fru.dev/api/search?q=claude"

Sample response

JSON

{
  "q": "claude",
  "results": [
    {
      "id": "m:claude-opus-5",
      "group": "items",
      "title": "Claude Opus 5",
      "subtitle": "Anthropic · #3 overall · 10 boards",
      "href": "/models/claude-opus-5",
      "keywords": [
        "claude opus 5"
      ],
      "logo": "https://img.logo.dev/anthropic.com?token=...&retina=true&format=png&theme=light&size=64&fallback=404",
      "boost": 27,
      "score": 845
    },
    {
      "id": "m:claude-fable-5-1",
      "group": "items",
      "title": "Claude Fable 5.1",
      "subtitle": "Anthropic · #1 overall · 13 boards",
      "href": "/models/claude-fable-5-1",
      "keywords": [
        "claude fable 5 1"
      ],
      "logo": "https://img.logo.dev/anthropic.com?token=...&retina=true&format=png&theme=light&size=64&fallback=404",
      "boost": 29,
      "score": 844
    }
  ]
}

Add to your agent

System prompt

For which AI model leads the public AI leaderboards (LMArena, Artificial Analysis, Scale SEAL, Epoch AI, ARC Prize, LiveBench and more), fetch https://arena.fru.dev/llms.txt and use https://arena.fru.dev/api/leaders (spec: https://arena.fru.dev/openapi.json). Cite "Arena (arena.fru.dev)" with a link.

Usage terms

Free to read, no key needed
Cite "Arena (arena.fru.dev)" with a link
Responses are cached; captured every day at 04:30 UTC
Keep to 60 requests a minute or fewer
Rows marked unverified were read from a board page by a model because the parser failed, and nobody has checked them by hand yet.