Every AI leaderboard

Leaderboards28Models ranked992

28 leaderboards · 992 models ranked

LMArena Document Arena what it measures, how it is scored and who leads it

Arenas · LMArena · Arena rating

Leaderboard

Scores from LMArena Document Arena, published Nov 9, 2026

RankModelArena rating
1Claude Fable 5Anthropic · as “Claude Fable 5 (High)”1,500
2Claude Fable 5.1Anthropic · as “Claude Fable 5.1 (Max)”1,498
3Claude Opus 4.6Anthropic · as “Claude Opus 4.6 (High)”1,497
4Claude Opus 4.7Anthropic · as “Claude Opus 4.7”1,494
5GPT-5.6 SolOpenAI · as “GPT-5.6 Sol (xHigh)”1,486
6Claude Opus 5Anthropic · as “Claude Opus 5 (High)”1,486
7Claude Opus 4.8Anthropic · as “Claude Opus 4.8 (High)”1,485
8GPT-5.5OpenAI · as “GPT-5.5 (High)”1,481
9Muse Spark 1.1Meta · as “Muse Spark 1.1”1,480
10Claude Sonnet 4.6Anthropic · as “Claude Sonnet 4.6”1,478
11GPT-5.6 TerraOpenAI · as “GPT-5.6 Terra (xHigh)”1,474
12Muse Spark 1.3Meta · as “Muse Spark 1.3 (Max)”1,474
13GPT-6 AstraOpenAI · as “GPT-6 Astra (Max)”1,474
14GPT-5.4OpenAI · as “GPT-5.4”1,471
15Gemini 3.8 FlashGoogle · as “Gemini 3.8 Flash (High)”1,469
16Claude Sonnet 5Anthropic · as “Claude Sonnet 5 (High)”1,469
17Gemini 3.7 FlashGoogle · as “Gemini 3.7 Flash (High)”1,468
18Claude Opus 4.5Anthropic · as “Claude Opus 4.5 (20251101)”1,467
19Muse SparkMeta · as “Muse Spark”1,467
20Grok 4.5SpaceXAI · as “Grok 4.5”1,466
21GPT-5.6 LunaOpenAI · as “GPT-5.6 Luna (xHigh)”1,460
22GPT-5.5 Instant (June 2026)OpenAI · as “GPT-5.5 Instant”1,460
23Gemini 3.1 Pro PreviewGoogle · as “Gemini 3.1 Pro Preview”1,459
24Gemini 3.5 FlashGoogle · as “Gemini 3.5 Flash (Medium)”1,459
25Gemini 3.6 FlashGoogle · as “Gemini 3.6 Flash (High)”1,454

About this benchmark

Task
Answer questions about uploaded documents and PDFs; people vote.
Dataset
Crowd-sourced documents, prompts and votes.
Method
Pairwise votes fitted with a Bradley-Terry model.
Metric
Arena rating
Organization
LMArena
Official leaderboard
arena.ai/leaderboard/document

Sources: each benchmark's paper (arXiv) and its official site or leaderboard. Scores are the boards captured here every day, exactly as published.