Benchmark Almanac

OpenDesign Arena

Kimi K3.1 Pro 91.0 leads since Oct 2026 Not in model cards

Models create design artifacts in the OpenDesign tool, and the board reports an evaluation score out of 100 with cost and speed per artifact.

Vendor board inside a design product. Kimi K3.1 Pro led with 91.0 in 2026-10. How scores are judged needs a check.

Made by OpenDesign Web and app front ends Tasks kept private Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Everyday Most adults, without training
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
Leaderboard open-design.ai
22
Entries on its own leaderboard, 19 models we track

Kimi K3.1 Pro leads with 91.0

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 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.

22 entries. Scroll sideways for the rest.

  1. 91.0 Kimi K3.1 Pro
  2. 90.3 GPT-6.1 Terra
  3. 90.1 GPT-6.1 Sol
  4. 89.9 GPT-6 Sol
  5. 86.8 Claude Sonnet 5.5
  6. 83.5 MiMo 2.6 Flash
  7. 83.1 GPT-6 Luna
  8. 82.9 Grok 4.7
  9. 82.7 GPT-6 Astra
  10. 82.2 Claude Haiku 5.5
  11. 81.7 MiniMax-M3.1-Flash
  12. 81.2 DeepSeek V4.1 Flash
  13. 81.2 Gemini 4 Argon
  14. 80.3 Claude Fable 5.1
  15. 80.1 GLM-5.3 Flash-X
  16. 77.6 GPT-5.6 Sol
  17. 76.2 Muse Wonder
  18. 74.6 Hunyuan H4 Preview
  19. 72.0 Qwen 3.8 Max
  20. 68.9 Gemini 3.8 Flash
  21. 66.6 Muse Spark 1.3
  22. 65.7 Kimi K3
  1. 1 Kimi K3.1 Pro 91.0
  2. 2 GPT-6.1 Terra 90.3
  3. 3 GPT-6.1 Sol 90.1
  4. 4 GPT-6 Sol 89.9
  5. 5 Claude Sonnet 5.5 86.8
  6. 6 MiMo 2.6 Flash 83.5
  7. 7 GPT-6 Luna 83.1
  8. 8 Grok 4.7 82.9
  9. 9 GPT-6 Astra 82.7
  10. 10 Claude Haiku 5.5 82.2
  11. 11 MiniMax-M3.1-Flash 81.7
  12. 12 DeepSeek V4.1 Flash 81.2
  13. 13 Gemini 4 Argon 81.2
  14. 14 Claude Fable 5.1 80.3
  15. 15 GLM-5.3 Flash-X 80.1
  16. 16 GPT-5.6 Sol 77.6
  17. 17 Muse Wonder 76.2
  18. 18 Hunyuan H4 Preview 74.6
  19. 19 Qwen 3.8 Max 72.0
  20. 20 Gemini 3.8 Flash 68.9
  21. 21 Muse Spark 1.3 66.6
  22. 22 Kimi K3 65.7
#EntryScoreSettingDateSource
1 Kimi K3.1 Pro 91.0 Oct 2026 Official leaderboard
2 GPT-6.1 Terra 90.3 Oct 2026 Official leaderboard
3 GPT-6.1 Sol 90.1 Sep 2026 Official leaderboard
4 GPT-6 Sol 89.9 Sep 2026 Official leaderboard
5 Claude Sonnet 5.5 86.8 Sep 2026 Official leaderboard
6 MiMo 2.6 Flash 83.5 Official leaderboard
7 GPT-6 Luna 83.1 Sep 2026 Official leaderboard
8 Grok 4.7 82.9 Sep 2026 Official leaderboard
9 GPT-6 Astra 82.7 Sep 2026 Official leaderboard
10 Claude Haiku 5.5 82.2 Oct 2026 Official leaderboard
11 MiniMax-M3.1-Flash 81.7 Oct 2026 Official leaderboard
12 DeepSeek V4.1 Flash 81.2 Sep 2026 Official leaderboard
13 Gemini 4 Argon 81.2 Sep 2026 Official leaderboard
14 Claude Fable 5.1 80.3 Sep 2026 Official leaderboard
15 GLM-5.3 Flash-X 80.1 Official leaderboard
16 GPT-5.6 Sol 77.6 Jul 2026 Official leaderboard
17 Muse Wonder 76.2 Oct 2026 Official leaderboard
18 Hunyuan H4 Preview 74.6 Official leaderboard
19 Qwen 3.8 Max 72.0 Aug 2026 Official leaderboard
20 Gemini 3.8 Flash 68.9 Sep 2026 Official leaderboard
21 Muse Spark 1.3 66.6 Sep 2026 Official leaderboard
22 Kimi K3 65.7 Jul 2026 Official leaderboard

The best score, 91.0, costs $1.46 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Terra comes within 0.7 of the best for $0.12 a task, 12 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Moonshot AI Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 100.0 $0.01 $0.10 $1.00 $10 Cost per task, log scale Kimi K3.1 Pro GPT-6.1 Terra GPT-6 Luna 0.00 20.0 40.0 60.0 80.0 100.0 $0.01 $0.10 $1.00 $10 Cost per task, log scale Kimi K3.1 Pro GPT-6.1 Terra
ModelEffortHarnessScoreCost per taskOutput tokens
Kimi K3.1 Pro 91.0 $1.46 32,679
GPT-6.1 Terra 90.3 $0.12 11,426
GPT-6.1 Sol 90.1 $0.38 23,311
GPT-6 Sol 89.9 $0.30 13,706
Claude Sonnet 5.5 86.8 $0.57 31,318
GPT-6 Luna 83.1 $0.02 17,265
Grok 4.7 82.9 $0.66 44,268
GPT-6 Astra 82.7 $1.61 15,544
Claude Haiku 5.5 82.2 $0.02 24,760
MiniMax-M3.1-Flash 81.7 $0.25 126,582
DeepSeek V4.1 Flash 81.2 $0.02 29,202
Gemini 4 Argon 81.2 $0.67 41,624
Claude Fable 5.1 80.3 $3.66 63,610
GPT-5.6 Sol 77.6 $0.54 14,521
Muse Wonder 76.2 $1.25 21,235
Qwen 3.8 Max 72.0 $0.60 28,629
Gemini 3.8 Flash 68.9 $0.21 27,851
Muse Spark 1.3 66.6 $0.27 13,428
Kimi K3 65.7 $0.61 16,066

7 models have held number one since 2023

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

OpenAI Anthropic xAI Moonshot AI
Own leaderboard 7 leaders
Kimi K3.1 Pro 91.0

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
60.0 70.0 80.0 90.0 100.0 2027 60.0 70.0 80.0 90.0 100.0 2027

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Coding

All 243 in Coding
  • SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Opus 5.5 93.9%
  • LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests. DeepSeek-V4-Pro 93.5%
  • DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
  • SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases. Seed2.1 Pro 59.8%
  • Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance. DeepSeek V4.1 Flash 3,471