Benchmark Almanac

DesignArena HTML Slides

GPT-6 Astra 1,535 leads since September 2026
Not in model cards

Agents write a self-contained HTML slide deck that is converted to PPTX. Users compare the decks and vote for the better one.

One of 40 DesignArena boards. DesignArena ranks entries with Bradley-Terry ratings from user votes.

Kind
Documents and spreadsheets
Level
Everyday
Scored
Ranked by votes
Answers
Open-ended
Tasks
All public
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
28
Models scored on its own leaderboard

GPT-6 Astra leads with 1,535

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

28 models. Scroll sideways for the rest. Bars start at 800.

  1. 1,535 GPT-6 Astra
  2. 1,489 GPT-6.1 Sol
  3. 1,477 Claude Opus 5.5
  4. 1,395 Claude Fable 5.1
  5. 1,181 GPT-5.6 Sol
  6. 1,173 Claude Fable 5
  7. 1,152 Kimi K3
  8. 1,144 Grok 4.6
  9. 1,135 Muse Wonder
  10. 1,116 Claude Opus 4.8
  11. 1,113 Claude Sonnet 5
  12. 1,100 Grok 4.5
  13. 1,096 Kimi K2.7 Code
  14. 1,090 Muse Spark 1.1
  15. 1,080 Kimi K2.6
  16. 1,077 MiniMax-M3
  17. 1,071 GLM-5.2
  18. 1,066 GLM-5.1
  19. 1,054 Qwen3.7-Max
  20. 1,048 Gemini 3.1 Pro
  21. 1,044 Gemini 3.6 Flash
  22. 1,043 GLM-5
  23. 1,038 Gemini 3.5 Flash
  24. 1,030 Muse Spark 1.2
  25. 1,016 GLM-5V-Turbo
  26. 1,002 GPT-5.6 Terra
  27. 963 GPT-5.5
  28. 905 Grok 4.3

Bars start at 800.

  1. 1 GPT-6 Astra xhigh, max 1,535
  2. 2 GPT-6.1 Sol max 1,489
  3. 3 Claude Opus 5.5 high 1,477
  4. 4 Claude Fable 5.1 1,395
  5. 5 GPT-5.6 Sol medium, xhigh 1,181
  6. 6 Claude Fable 5 1,173
  7. 7 Kimi K3 1,152
  8. 8 Grok 4.6 1,144
  9. 9 Muse Wonder max 1,135
  10. 10 Claude Opus 4.8 1,116
  11. 11 Claude Sonnet 5 1,113
  12. 12 Grok 4.5 1,100
  13. 13 Kimi K2.7 Code 1,096
  14. 14 Muse Spark 1.1 1,090
  15. 15 Kimi K2.6 1,080
  16. 16 MiniMax-M3 1,077
  17. 17 GLM-5.2 1,071
  18. 18 GLM-5.1 1,066
  19. 19 Qwen3.7-Max 1,054
  20. 20 Gemini 3.1 Pro 1,048
  21. 21 Gemini 3.6 Flash 1,044
  22. 22 GLM-5 1,043
  23. 23 Gemini 3.5 Flash 1,038
  24. 24 Muse Spark 1.2 1,030
  25. 25 GLM-5V-Turbo 1,016
  26. 26 GPT-5.6 Terra 1,002
  27. 27 GPT-5.5 963
  28. 28 Grok 4.3 905

Source: Official DesignArena HTML Slides leaderboard

#ModelScoreSettingDateSource
1 GPT-6 Astra 1,535 xhigh, max Sep 2026 Official leaderboard
2 GPT-6.1 Sol 1,489 max Sep 2026 Official leaderboard
3 Claude Opus 5.5 1,477 high Sep 2026 Official leaderboard
4 Claude Fable 5.1 1,395 Sep 2026 Official leaderboard
5 GPT-5.6 Sol 1,181 medium, xhigh Jul 2026 Official leaderboard
6 Claude Fable 5 1,173 Jun 2026 Official leaderboard
7 Kimi K3 1,152 Jul 2026 Official leaderboard
8 Grok 4.6 1,144 Aug 2026 Official leaderboard
9 Muse Wonder 1,135 max Oct 2026 Official leaderboard
10 Claude Opus 4.8 1,116 May 2026 Official leaderboard
11 Claude Sonnet 5 1,113 Jun 2026 Official leaderboard
12 Grok 4.5 1,100 Jul 2026 Official leaderboard
13 Kimi K2.7 Code 1,096 Jun 2026 Official leaderboard
14 Muse Spark 1.1 1,090 Jul 2026 Official leaderboard
15 Kimi K2.6 1,080 Apr 2026 Official leaderboard
16 MiniMax-M3 1,077 Jun 2026 Official leaderboard
17 GLM-5.2 1,071 Jun 2026 Official leaderboard
18 GLM-5.1 1,066 Apr 2026 Official leaderboard
19 Qwen3.7-Max 1,054 May 2026 Official leaderboard
20 Gemini 3.1 Pro 1,048 Feb 2026 Official leaderboard
21 Gemini 3.6 Flash 1,044 Jul 2026 Official leaderboard
22 GLM-5 1,043 Feb 2026 Official leaderboard
23 Gemini 3.5 Flash 1,038 May 2026 Official leaderboard
24 Muse Spark 1.2 1,030 Aug 2026 Official leaderboard
25 GLM-5V-Turbo 1,016 Apr 2026 Official leaderboard
26 GPT-5.6 Terra 1,002 Jul 2026 Official leaderboard
27 GPT-5.5 963 Apr 2026 Official leaderboard
28 Grok 4.3 905 Apr 2026 Official leaderboard

9 models have held number one on the official leaderboard since 2023

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

Z.ai Google DeepMind Moonshot AI Anthropic OpenAI
Own leaderboard 9 leaders
GPT-6 Astra 1,535

Source: Official DesignArena HTML Slides 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
800 1,000 1,200 1,400 1,600 2026 800 1,000 1,200 1,400 1,600 2026

Source: Official DesignArena HTML Slides 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 Economic work

All 97 in Economic work
  • JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
  • GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate. Seed2.1 Pro 87.9%
  • GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
  • AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
  • CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Claude Fable 5 75.9%
  • OfficeQA In use Grounded reasoning over a large corpus of U.S. Treasury Bulletin documents: locate tables and compute precise numeric answers. Claude Opus 4.7 86.3%