Landmark Arena
We show this table for reference; we do not rank on it.
Kaggle's Landmark Arena has models build 20 famous landmarks, including the Seven Wonders of the Ancient World, as single-page three.js web apps, and people vote for the better scene.
Evaluation results published by Kaggle. How Kaggle Benchmarks reruns tests.
Arena rating on Landmark Arena v1 — Latest evaluation November 13, 2026
We mirror the published arena rating view for Landmark Arena v1. Claude Opus 5.5 has the highest reported value at 1792. A rating gap smaller than the two rows' intervals is not a reliable difference. We do not use these results to rank models overall.
Claude Opus 5.5
Max Reasoning; 146 votes; ±117; Evaluated 2026-11-13
GPT-6 Astra
Max Reasoning; 110 votes; ±120; Evaluated 2026-11-13
GPT-6.1 Sol
Max Reasoning; 162 votes; ±110; Evaluated 2026-11-12
14 source rowsCross-domainCurrentDisplay onlyUpdated Latest evaluation November 13, 2026
Arena rating table (14 source rows)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5.5Creator not reportedMax Reasoning; 146 votes; ±117; Evaluated 2026-11-13 | 1792 | Not reported | Unknown |
| 2 | GPT-6 AstraCreator not reportedMax Reasoning; 110 votes; ±120; Evaluated 2026-11-13 | 1755 | Not reported | Unknown |
| 3 | GPT-6.1 SolCreator not reportedMax Reasoning; 162 votes; ±110; Evaluated 2026-11-12 | 1694 | Not reported | Unknown |
| 4 | Claude Fable 5.1Creator not reportedMax Reasoning; 167 votes; ±108; Evaluated 2026-11-12 | 1607 | Not reported | Unknown |
| 5 | Gemini 4 ArgonCreator not reportedHigh Reasoning; 169 votes; ±107; Evaluated 2026-11-12 | 1595 | Not reported | Unknown |
| 6 | Gemini 4 CarbonCreator not reportedHigh Reasoning; 67 votes; ±124; Evaluated 2026-11-13 | 1556 | Not reported | Unknown |
| 7 | GLM-5.3Creator not reportedMax Reasoning; 138 votes; ±109; Evaluated 2026-11-13 | 1484 | Not reported | Unknown |
| 8 | Qwen 3.8 MaxCreator not reportedXhigh Reasoning; 121 votes; ±111; Evaluated 2026-11-13 | 1475 | Not reported | Unknown |
| 9 | Deepseek V4.1 FlashCreator not reportedHigh Reasoning; 181 votes; ±105; Evaluated 2026-11-13 | 1431 | Not reported | Unknown |
| 10 | Grok 4.7Creator not reportedXhigh Reasoning; 171 votes; ±107; Evaluated 2026-11-12 | 1416 | Not reported | Unknown |
| 11 | Gemini 3.8 FlashCreator not reportedHigh Reasoning; 181 votes; ±106; Evaluated 2026-11-13 | 1368 | Not reported | Unknown |
| 12 | Kimi K3Creator not reportedMax Reasoning; 27 votes; ±157; Evaluated 2026-11-12 | 1357 | Not reported | Unknown |
| 13 | Grok 4.6Creator not reportedXhigh Reasoning; 155 votes; ±110; Evaluated 2026-11-13 | 1308 | Not reported | Unknown |
| 14 | Claude Sonnet 5Creator not reported189 votes; ±114; Evaluated 2026-11-13 | 1162 | Not reported | Unknown |
How to read this leaderboard
Read the rating as a preference ranking for 3D landmark scenes built in three.js among the models on this board, with its vote count and interval. It does not measure accuracy.
Operator receipt: 14 sourced rows are currently displayable on this page; the highest published score among these 14 source rows is Claude Opus 5.5 at 1792, which does not establish a market leader.
Honest limit: The rating moves as votes arrive, and Kaggle publishes no confidence level for its interval. Rows with few votes have wide intervals, so neighbouring models are often not separable. Voters are self-selected Kaggle visitors. The board declares no license; we keep published aggregates with attribution and no task or output contents.
How to read Landmark Arena
A page qualifies only if it is self-contained and renders. It tests 3D scene building in browser code. Kaggle shows two anonymous outputs for the same prompt and a person picks the better one. The Arena column is the relative rating Kaggle fits to those votes, published with an interval; it is not a percentage.
The rating moves as votes arrive, and Kaggle publishes no confidence level for its interval. Rows with few votes have wide intervals, so neighbouring models are often not separable. Voters are self-selected Kaggle visitors. The board declares no license; we keep published aggregates with attribution and no task or output contents.
Snapshot
A rating gap smaller than the two rows' intervals is not a reliable difference. The table retains each source configuration and its reported value.
14 source rows are shown for Landmark Arena. The benchmark falls in the Cross-domain category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Landmark Arena is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Landmark Arena
Year
2026
Tasks
20 famous landmarks built as three.js web pages, rated by human votes
Format
Relative rating fitted to human pairwise votes
Difficulty
3D scenes in browser code
A page qualifies only if it is self-contained and renders. It tests 3D scene building in browser code. Kaggle shows two anonymous outputs for the same prompt and a person picks the better one. The Arena column is the relative rating Kaggle fits to those votes, published with an interval; it is not a percentage. These results are display-only and do not enter overall or category rankings.
Freshness and provenance
Version
Landmark Arena 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does Landmark Arena measure?
Kaggle's Landmark Arena has models build 20 famous landmarks, including the Seven Wonders of the Ancient World, as single-page three.js web apps, and people vote for the better scene.
Which source row has the highest reported Landmark Arena value?
Claude Opus 5.5 has the highest reported Landmark Arena value at 1792. A rating gap smaller than the two rows' intervals is not a reliable difference. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on Landmark Arena?
The Latest evaluation November 13, 2026 snapshot contains 14 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.