Benchmark Almanac

Document Arena

Claude Fable 5 1,500 leads since June 2026
Not in model cards

People upload their own PDFs with a question, compare answers from two anonymous models, and vote for the better one. Votes become Bradley-Terry ratings.

PDF uploads opened in February 2026 and the ranked board went live on March 3rd, 2026. It had 44 models and 404,715 votes on the September 13th, 2026 update. A style-controlled view also exists.

Released
March 2026
Kind
Reasoning over long documents
Level
Everyday
Scored
Ranked by votes
Answers
Open-ended
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
40
Entries on its own leaderboard, 39 models we track

Claude Fable 5 leads with 1,500 among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

40 entries. Scroll sideways for the rest. Bars start at 1,400.

  1. 1,500 Claude Fable 5
  2. 1,498 Claude Fable 5.1
  3. 1,497 Claude Opus 4.6
  4. 1,494 Claude Opus 4.7
  5. 1,486 GPT-5.6 Sol
  6. 1,486 Claude Opus 5
  7. 1,485 Claude Opus 4.8
  8. 1,481 GPT-5.5
  9. 1,480 Muse Spark 1.1
  10. 1,478 Claude Sonnet 4.6
  11. 1,474 GPT-5.6 Terra
  12. 1,474 Muse Spark 1.3
  13. 1,474 GPT-6 Astra
  14. 1,471 GPT-5.4
  15. 1,469 Gemini 3.8 Flash
  16. 1,469 Claude Sonnet 5
  17. 1,468 Gemini 3.7 Flash
  18. 1,468 Claude Opus 4.5
  19. 1,467 Muse Spark
  20. 1,466 Grok 4.5
  21. 1,460 GPT-5.6 Luna
  22. 1,460 GPT-5.5 Instant
  23. 1,459 Gemini 3.1 Pro
  24. 1,459 Gemini 3.5 Flash
  25. 1,454 Gemini 3.6 Flash
  26. 1,452 Gemma 4 31B IT
  27. 1,451 Gemini 3 Pro
  28. 1,450 Kimi K2.6
  29. 1,450 Grok 4.6
  30. 1,448 Claude Sonnet 4.5
  31. 1,446 Qwen3.7-Plus
  32. 1,440 Grok 4.20 Beta (Reasoning, 0309)
  33. 1,440 Gemini 3 Flash
  34. 1,438 Kimi K2.5
  35. 1,435 MiniMax-M3
  36. 1,433 Gemini 2.5 Pro
  37. 1,425 GPT-5.2
  38. 1,419 Claude Haiku 4.5
  39. 1,416 GPT-5.1
  40. 1,406 GLM-5V-Turbo

Bars start at 1,400.

  1. 1 Claude Fable 5 high 1,500
  2. 2 Claude Fable 5.1 max 1,498
  3. 3 Claude Opus 4.6 high 1,497
  4. 4 Claude Opus 4.7 1,494
  5. 5 GPT-5.6 Sol xhigh 1,486
  6. 6 Claude Opus 5 high 1,486
  7. 7 Claude Opus 4.8 high 1,485
  8. 8 GPT-5.5 high 1,481
  9. 9 Muse Spark 1.1 1,480
  10. 10 Claude Sonnet 4.6 1,478
  11. 11 GPT-5.6 Terra xhigh 1,474
  12. 12 Muse Spark 1.3 max 1,474
  13. 13 GPT-6 Astra max 1,474
  14. 14 GPT-5.4 1,471
  15. 15 Gemini 3.8 Flash high 1,469
  16. 16 Claude Sonnet 5 high 1,469
  17. 17 Gemini 3.7 Flash high 1,468
  18. 18 Claude Opus 4.5 1,468
  19. 19 Muse Spark 1,467
  20. 20 Grok 4.5 1,466
  21. 21 GPT-5.6 Luna xhigh 1,460
  22. 22 GPT-5.5 Instant 1,460
  23. 23 Gemini 3.1 Pro 1,459
  24. 24 Gemini 3.5 Flash medium 1,459
  25. 25 Gemini 3.6 Flash high 1,454
  26. 26 Gemma 4 31B IT 1,452
  27. 27 Gemini 3 Pro 1,451
  28. 28 Kimi K2.6 1,450
  29. 29 Grok 4.6 high 1,450
  30. 30 Claude Sonnet 4.5 1,448
  31. 31 Qwen3.7-Plus 1,446
  32. 32 Grok 4.20 Beta (Reasoning, 0309) 1,440
  33. 33 Gemini 3 Flash 1,440
  34. 34 Kimi K2.5 thinking 1,438
  35. 35 MiniMax-M3 1,435
  36. 36 Gemini 2.5 Pro Jun 2025 1,433
  37. 37 GPT-5.2 1,425
  38. 38 Claude Haiku 4.5 1,419
  39. 39 GPT-5.1 1,416
  40. 40 GLM-5V-Turbo 1,406

Source: Official Document Arena leaderboard

#EntryScoreSettingDateSource
1 Claude Fable 5 1,500 high Jun 2026 Official leaderboard
2 Claude Fable 5.1 1,498 max Sep 2026 Official leaderboard
3 Claude Opus 4.6 1,497 high Feb 2026 Official leaderboard
4 Claude Opus 4.7 1,494 Apr 2026 Official leaderboard
5 GPT-5.6 Sol 1,486 xhigh Jul 2026 Official leaderboard
6 Claude Opus 5 1,486 high Jul 2026 Official leaderboard
7 Claude Opus 4.8 1,485 high May 2026 Official leaderboard
8 GPT-5.5 1,481 high Apr 2026 Official leaderboard
9 Muse Spark 1.1 1,480 Jul 2026 Official leaderboard
10 Claude Sonnet 4.6 1,478 Feb 2026 Official leaderboard
11 GPT-5.6 Terra 1,474 xhigh Jul 2026 Official leaderboard
12 Muse Spark 1.3 1,474 max Sep 2026 Official leaderboard
13 GPT-6 Astra 1,474 max Sep 2026 Official leaderboard
14 GPT-5.4 1,471 Mar 2026 Official leaderboard
15 Gemini 3.8 Flash 1,469 high Sep 2026 Official leaderboard
16 Claude Sonnet 5 1,469 high Jun 2026 Official leaderboard
17 Gemini 3.7 Flash 1,468 high Aug 2026 Official leaderboard
18 Claude Opus 4.5 1,468 Nov 2025 Official leaderboard
19 Muse Spark 1,467 Apr 2026 Official leaderboard
20 Grok 4.5 1,466 Jul 2026 Official leaderboard
21 GPT-5.6 Luna 1,460 xhigh Jul 2026 Official leaderboard
22 GPT-5.5 Instant 1,460 May 2026 Official leaderboard
23 Gemini 3.1 Pro 1,459 Feb 2026 Official leaderboard
24 Gemini 3.5 Flash 1,459 medium May 2026 Official leaderboard
25 Gemini 3.6 Flash 1,454 high Jul 2026 Official leaderboard
26 Gemma 4 31B IT 1,452 Apr 2026 Official leaderboard
27 Gemini 3 Pro 1,451 Nov 2025 Official leaderboard
28 Kimi K2.6 1,450 Apr 2026 Official leaderboard
29 Grok 4.6 1,450 high Aug 2026 Official leaderboard
30 Claude Sonnet 4.5 1,448 Sep 2025 Official leaderboard
31 Qwen3.7-Plus 1,446 Jun 2026 Official leaderboard
32 Grok 4.20 Beta (Reasoning, 0309) 1,440 Nov 2026 Official leaderboard
33 Gemini 3 Flash 1,440 Dec 2025 Official leaderboard
34 Kimi K2.5 1,438 thinking Jan 2026 Official leaderboard
35 MiniMax-M3 1,435 Jun 2026 Official leaderboard
36 Gemini 2.5 Pro Jun 2025 1,433 Jun 2025 Official leaderboard
37 GPT-5.2 1,425 Dec 2025 Official leaderboard
38 Claude Haiku 4.5 1,419 Oct 2025 Official leaderboard
39 GPT-5.1 1,416 Nov 2025 Official leaderboard
40 GLM-5V-Turbo 1,406 Apr 2026 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude Opus 4.6
high
1,497
Claude Opus 4.7
high
1,494
Claude Opus 4.8
high
1,485
GPT-5.5
high
1,481
Gemini 3.5 Flash
medium
1,459
GPT-5.2
high
1,425

Source: Official Document Arena leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5 high 1,500
Claude Fable 5.1 max 1,498
Claude Opus 4.6 high 1,497
Claude Opus 4.6 1,496
Claude Opus 4.7 1,494
Claude Opus 4.7 high 1,494
GPT-5.6 Sol xhigh 1,486
Claude Opus 5 high 1,486
Claude Opus 4.8 high 1,485
GPT-5.5 high 1,481
Muse Spark 1.1 1,480
GPT-5.5 1,479
Claude Sonnet 4.6 1,478
Claude Opus 4.8 1,476
GPT-5.6 Terra xhigh 1,474
Muse Spark 1.3 max 1,474
GPT-6 Astra max 1,474
GPT-5.4 1,471
Gemini 3.8 Flash high 1,469
Claude Sonnet 5 high 1,469
Gemini 3.7 Flash high 1,468
Claude Opus 4.5 1,468
Muse Spark 1,467
Grok 4.5 1,466
GPT-5.6 Luna xhigh 1,460
GPT-5.5 Instant 1,460
Gemini 3.1 Pro 1,459
Gemini 3.5 Flash medium 1,459
Gemini 3.5 Flash high 1,455
Gemini 3.6 Flash high 1,454
Gemma 4 31B IT 1,452
Gemini 3 Pro 1,451
Kimi K2.6 1,450
Grok 4.6 high 1,450
Claude Sonnet 4.5 1,448
Qwen3.7-Plus 1,446
Gemini 3 Flash 1,440
Kimi K2.5 thinking 1,438
MiniMax-M3 1,435
Gemini 2.5 Pro Jun 2025 1,433
GPT-5.2 1,425
GPT-5.2 high 1,422
Claude Haiku 4.5 1,419
GPT-5.1 1,416
GLM-5V-Turbo 1,406

2 models have held number one on the official leaderboard since its release in March 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Own leaderboard Not tested since Sep 2026
Claude Fable 5 1,500

Source: Official Document Arena leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
1,400 1,425 1,450 1,475 1,500 2025 2026 1,400 1,425 1,450 1,475 1,500 2025 2026

Source: Official Document Arena leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Long context

All 79 in Long context
  • MRCR v2 In use Multi-round coreference: retrieve the n-th of several similar requests hidden in a long synthetic conversation, up to 1M tokens. Muse Spark 1.3 98.5%
  • LongBench v2 In use 503 multiple-choice questions over 8K-2M-word contexts (QA, dialogue, code repos, structured data). Accuracy. Inkling 2 Small 70.6%
  • GraphWalks In use A graph is given as an edge list in long context. The model answers BFS or parent queries, scored by F1. Claude Fable 5 99.96%
  • OpenAI MRCR v2 In use Models read a long synthetic chat that hides 2, 4 or 8 identical writing requests and must reproduce the response to the i-th one. Scored by mean string match ratio per context bin. Gemini 3.7 Flash 89%
  • CL-bench In use Model must learn new knowledge from a supplied long context (rules, procedures, data) and solve tasks. Strict LLM-verified rubrics. GPT-5.4 27.9%
  • AA-LCR 1.1 In use Models answer 100 questions that need reasoning across sets of real documents of about 100K tokens. An LLM equality checker with a fixed system prompt compares each answer with the key. Step 5 Preview 88.3%