Skip to main content
BenchLM
Data

Game Arena

We show this table for reference; we do not rank on it.

Data verified

Kaggle Game Arena fits one Bradley-Terry rating to results across strategic games. Matches played and games covered accompany the rating; it is not a percentage or a chess Elo rating.

Evaluation results published by Kaggle. How Kaggle Benchmarks reruns tests.

Kaggle labels this benchmark Apache 2.0. We retain aggregate measurements with attribution and source links. That declaration does not establish rights to third-party documents, videos, or other underlying task material; no task contents are included. Original article terms.

Pooled Bradley-Terry rating on Game Arena — October 19, 2026

We mirror the published pooled bradley-terry rating view for Game Arena. Gemini 4 Argon has the highest reported value at 535. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. We do not use these results to rank models overall.

1

Gemini 4 Argon

Kaggle benchmark v1; task revision 1; Evaluated 2026-10-19

535
2

GPT-6 Astra

Kaggle benchmark v1; task revision 1; Evaluated 2026-10-19

491
3

Claude Opus 5

Kaggle benchmark v1; task revision 1; Evaluated 2026-10-19

436

39 source rowsReasoningCurrentDisplay onlyUpdated October 19, 2026

Pooled Bradley-Terry rating table (39 source rows)

Pooled Bradley-Terry rating results for Game Arena
RankModel / configurationScoreParameters (B)Open / closed
1Gemini 4 ArgonCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
535
Not reportedUnknown
2GPT-6 AstraCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
491
Not reportedUnknown
3Claude Opus 5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
436
Not reportedUnknown
4GPT-6.1 SolCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
436
Not reportedUnknown
5Gemini 3.8 FlashCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
429
Not reportedUnknown
6Claude Fable 5.1Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
418
Not reportedUnknown
7GPT-5.5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
407
Not reportedUnknown
8Claude Opus 5.5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
398
Not reportedUnknown
9Gemini 3.1 Pro PreviewCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
381
Not reportedUnknown
10Gemini 3.7 FlashCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
368
Not reportedUnknown
11GPT-5.6 SolCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
352
Not reportedUnknown
12GPT-5.4Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
352
Not reportedUnknown
13Grok 4.7Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
335
Not reportedUnknown
14GPT-6 SolCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
331
Not reportedUnknown
15Gemini 3.6 FlashCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
325
Not reportedUnknown
16Claude Sonnet 5.5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
324
Not reportedUnknown
17Claude Opus 4.8Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
321
Not reportedUnknown
18Gemini 3.5 FlashCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
321
Not reportedUnknown
19Grok 4.6Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
316
Not reportedUnknown
20GPT-5.2Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
283
Not reportedUnknown
21Gemini 3 Flash PreviewCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
278
Not reportedUnknown
22o3Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
268
Not reportedUnknown
23Claude Opus 4.6Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
259
Not reportedUnknown
24Grok 4.5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
258
Not reportedUnknown
25Claude Sonnet 5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
256
Not reportedUnknown
26Claude Haiku 5.5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
251
Not reportedUnknown
27Claude Opus 4.7Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
209
Not reportedUnknown
28Claude Sonnet 4.6Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
200
Not reportedUnknown
29GPT-5.6 TerraCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
195
Not reportedUnknown
30Grok 4.20 ReasoningCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
189
Not reportedUnknown
31Grok 4Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
181
Not reportedUnknown
32Claude Sonnet 4.5Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
173
Not reportedUnknown
33GPT-6 LunaCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
169
Not reportedUnknown
34GPT-5.6 LunaCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
164
Not reportedUnknown
35GPT-5 miniCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
154
Not reportedUnknown
36Grok 4.3Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
145
Not reportedUnknown
37GPT-5.4 miniCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
140
Not reportedUnknown
38Grok 4.1 Fast ReasoningCreator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
135
Not reportedUnknown
39DeepSeek V3.2Creator not reportedKaggle benchmark v1; task revision 1; Evaluated 2026-10-19
0
Not reportedUnknown

How to read Game Arena

Game Arena fits a pooled Bradley-Terry model to win, loss, and draw outcomes across its games. Each game receives equal total weight, and only the core variant counts when a game has several formats. The published Score is a relative rating, with matches played and games covered retained for each model.

The rating depends on this opponent pool and game mix. It is neither a percentage nor a chess Elo rating, and a model with fewer covered games has a different evidence base. Published interval widths remain in the source measurements without an inferred confidence level.

The complete official download includes models hidden by the source page's default filters. Source names and task versions stay separate from canonical model profiles; these results do not enter overall or category rankings.

Snapshot

39 source rowsBenchmark v1Complete downloadDisplay only
Official Kaggle leaderboardOfficial leaderboard JSONEvaluation methodBenchmark-page Apache 2.0 declaration

These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. The table retains each source configuration and its reported value.

39 source rows are shown for Game Arena. The benchmark falls in the Reasoning category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Game Arena is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Game Arena

Year

2025

Tasks

Strategic games; per-model coverage retained

Format

Pooled pairwise game outcomes

Game Arena fits a pooled Bradley-Terry model to win, loss, and draw outcomes across its games. Each game receives equal total weight, and only the core variant counts when a game has several formats. The published Score is a relative rating, with matches played and games covered retained for each model. The rating depends on this opponent pool and game mix. It is neither a percentage nor a chess Elo rating, and a model with fewer covered games has a different evidence base. Published interval widths remain in the source measurements without an inferred confidence level. The complete official download includes models hidden by the source page's default filters. Source names and task versions stay separate from canonical model profiles; these results do not enter overall or category rankings.

Game Arena methodologyPublic benchmark source

Freshness and provenance

Version

Game Arena 2025

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

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 Game Arena measure?

Kaggle Game Arena fits one Bradley-Terry rating to results across strategic games. Matches played and games covered accompany the rating; it is not a percentage or a chess Elo rating.

Which source row has the highest reported Game Arena value?

Gemini 4 Argon has the highest reported Game Arena value at 535. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Game Arena?

The October 19, 2026 snapshot contains 39 AI models.

Last updated: October 19, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 6,100+ subscribers.

One email each week. Unsubscribe anytime.