Benchmark Almanac

Game Arena Four-in-a-Row (Generality)

GPT-5 92.0 leads since August 2025
Not in model cards

Models place pieces on any empty square of a four by nine grid. Four in a line wins. The tournaments use different board encodings.

Generality ranks 27 configurations of 23 models in each tournament. The ASCII grid and the compact position notation each have 1,404 games.

Made by
Generality, Inc.
Kind
Multi-agent games
Level
Everyday
Scored
Ranked by matches
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Actions
Interaction
Multi-agent
Environment
Four-in-a-row
Items from
Live play or votes
Refresh
Live
Run by
A host
In views
Games, Agents, Kept fresh
Strongest model tested
GPT-5 #71 on the capability index
Last model added
November 2025
Scores refreshed
November 20th, 2026
23
Entries on its own leaderboard, 22 models we track

GPT-5 leads with 92.0 among models released by November 2025 GPT-5 leads with 92.0 among models released by November 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. This view shows only models we track, each at its best.

The newest model with a score here came out in November 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-5 ASCII board, reasoning medium 92.0
  2. 2 gpt-oss-120b ASCII board, reasoning high 89.0
  3. 3 o3 ASCII board, reasoning high 79.5
  4. 4 Grok 4 ASCII board, thinking enabled 79.0
  5. 5 Claude Sonnet 4.5 ASCII board, thinking budget 31000 71.5
  6. 6 DeepSeek-V3.2-Exp FEN board, thinking enabled 71.5
  7. 7 GPT-5 mini FEN board, reasoning high 70.0
  8. 8 Claude Sonnet 4 FEN board, thinking budget 31000 69.5
  9. 9 Gemini 2.5 Pro Jun 2025, ASCII board, thinking budget -1 67.0
  10. 10 Qwen3-235B-A22B-Thinking Jul 2025, ASCII board, reasoning medium, thinking enabled 67.0
  11. 11 Claude Opus 4.1 ASCII board, thinking budget 31000 66.0
  12. 12 qwen/qwen3-next-80b-a3b-thinking-maas ASCII board, thinking enabled 61.0
  13. 13 DeepSeek-R1 May 2025, FEN board, thinking enabled 61.0
  14. 14 DeepSeek-V3.1 FEN board, thinking enabled 58.0
  15. 15 Grok 4 Fast FEN board, thinking enabled 50.5
  16. 16 Kimi K2 Thinking ASCII board, thinking enabled 50.0
  17. 17 Qwen3-Max FEN board, thinking non-reasoning model 40.0
  18. 18 GLM-4.6 FEN board, thinking enabled 36.5
  19. 19 grok-code-fast-1 ASCII board, thinking enabled 31.0
  20. 20 Llama 3.3 70B Instruct ASCII board, thinking non-reasoning model 26.0
  21. 21 Kimi K2 Sep 2025, FEN board, thinking non-reasoning model 21.0
  22. 22 GLM-4.5 FEN board, reasoning medium, thinking enabled 17.0
  23. 23 GPT-4o Aug 2024, ASCII board, reasoning non-reasoning model 13.0

Source: Official Game Arena Four-in-a-Row (Generality) leaderboard

Source: Official Game Arena Four-in-a-Row (Generality) leaderboard

#EntryScoreSettingDateSource
1 GPT-5 92.0 ASCII board, reasoning medium Aug 2025 Official leaderboard
2 gpt-oss-120b 89.0 ASCII board, reasoning high Aug 2025 Official leaderboard
3 o3 79.5 ASCII board, reasoning high Apr 2025 Official leaderboard
4 Grok 4 79.0 ASCII board, thinking enabled Jul 2025 Official leaderboard
5 Claude Sonnet 4.5 71.5 ASCII board, thinking budget 31000 Sep 2025 Official leaderboard
6 DeepSeek-V3.2-Exp 71.5 FEN board, thinking enabled Sep 2025 Official leaderboard
7 GPT-5 mini 70.0 FEN board, reasoning high Aug 2025 Official leaderboard
8 Claude Sonnet 4 69.5 FEN board, thinking budget 31000 May 2025 Official leaderboard
9 Gemini 2.5 Pro Jun 2025 67.0 ASCII board, thinking budget -1 Jun 2025 Official leaderboard
10 Qwen3-235B-A22B-Thinking Jul 2025 67.0 ASCII board, reasoning medium, thinking enabled Jul 2025 Official leaderboard
11 Claude Opus 4.1 66.0 ASCII board, thinking budget 31000 Aug 2025 Official leaderboard
12 qwen/qwen3-next-80b-a3b-thinking-maas 61.0 ASCII board, thinking enabled Official leaderboard
13 DeepSeek-R1 May 2025 61.0 FEN board, thinking enabled May 2025 Official leaderboard
14 DeepSeek-V3.1 58.0 FEN board, thinking enabled Aug 2025 Official leaderboard
15 Grok 4 Fast 50.5 FEN board, thinking enabled Sep 2025 Official leaderboard
16 Kimi K2 Thinking 50.0 ASCII board, thinking enabled Nov 2025 Official leaderboard
17 Qwen3-Max 40.0 FEN board, thinking non-reasoning model Sep 2025 Official leaderboard
18 GLM-4.6 36.5 FEN board, thinking enabled Sep 2025 Official leaderboard
19 grok-code-fast-1 31.0 ASCII board, thinking enabled Aug 2025 Official leaderboard
20 Llama 3.3 70B Instruct 26.0 ASCII board, thinking non-reasoning model Dec 2024 Official leaderboard
21 Kimi K2 Sep 2025 21.0 FEN board, thinking non-reasoning model Sep 2025 Official leaderboard
22 GLM-4.5 17.0 FEN board, reasoning medium, thinking enabled Aug 2025 Official leaderboard
23 GPT-4o Aug 2024 13.0 ASCII board, reasoning non-reasoning model Aug 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5 92.0 ASCII board, reasoning medium Aug 2025 Official leaderboard
2 gpt-oss-120b 89.0 ASCII board, reasoning high Aug 2025 Official leaderboard
3 o3 79.5 ASCII board, reasoning high Apr 2025 Official leaderboard
4 Grok 4 79.0 ASCII board, thinking enabled Jul 2025 Official leaderboard
5 Claude Sonnet 4.5 71.5 ASCII board, thinking budget 31000 Sep 2025 Official leaderboard
6 DeepSeek-V3.2-Exp 71.5 FEN board, thinking enabled Sep 2025 Official leaderboard
7 GPT-5 mini 70.0 FEN board, reasoning high Aug 2025 Official leaderboard
8 Claude Sonnet 4 69.5 FEN board, thinking budget 31000 May 2025 Official leaderboard
9 Gemini 2.5 Pro Jun 2025 67.0 ASCII board, thinking budget -1 Jun 2025 Official leaderboard
10 Qwen3-235B-A22B-Thinking Jul 2025 67.0 ASCII board, reasoning medium, thinking enabled Jul 2025 Official leaderboard
11 Claude Opus 4.1 66.0 ASCII board, thinking budget 31000 Aug 2025 Official leaderboard
12 DeepSeek-R1 May 2025 61.0 FEN board, thinking enabled May 2025 Official leaderboard
13 DeepSeek-V3.1 58.0 FEN board, thinking enabled Aug 2025 Official leaderboard
14 Grok 4 Fast 50.5 FEN board, thinking enabled Sep 2025 Official leaderboard
15 Kimi K2 Thinking 50.0 ASCII board, thinking enabled Nov 2025 Official leaderboard
16 Qwen3-Max 40.0 FEN board, thinking non-reasoning model Sep 2025 Official leaderboard
17 GLM-4.6 36.5 FEN board, thinking enabled Sep 2025 Official leaderboard
18 grok-code-fast-1 31.0 ASCII board, thinking enabled Aug 2025 Official leaderboard
19 Llama 3.3 70B Instruct 26.0 ASCII board, thinking non-reasoning model Dec 2024 Official leaderboard
20 Kimi K2 Sep 2025 21.0 FEN board, thinking non-reasoning model Sep 2025 Official leaderboard
21 GLM-4.5 17.0 FEN board, reasoning medium, thinking enabled Aug 2025 Official leaderboard
22 GPT-4o Aug 2024 13.0 ASCII board, reasoning non-reasoning model Aug 2024 Official leaderboard

More effort adds a median 10 score

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
GPT-5
FEN board, … ASCII board…
92.0
gpt-oss-120b
FEN board, … ASCII board… ASCII board…
89.0
o3
FEN board, … ASCII board…
79.5
Grok 4
FEN board, … ASCII board…
79.0
Claude Sonnet 4.5
FEN board, … ASCII board…
71.5
DeepSeek-V3.2-Exp
ASCII board… FEN board, …
71.5
GPT-5 mini
ASCII board… FEN board, … FEN board, …
70.0
Claude Sonnet 4
ASCII board… FEN board, …
69.5
Gemini 2.5 Pro
FEN board, … ASCII board…
67.0
Qwen3-235B-A22B-Thinking
FEN board, … ASCII board…
67.0
Claude Opus 4.1
FEN board, … ASCII board…
66.0
DeepSeek-R1
ASCII board… FEN board, …
61.0

Source: Official Game Arena Four-in-a-Row (Generality) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5 ASCII board, reasoning medium 92.0
GPT-5 ASCII board, reasoning high 91.0
gpt-oss-120b ASCII board, reasoning high 89.0
GPT-5 FEN board, reasoning high 88.0
GPT-5 FEN board, reasoning medium 81.0
o3 ASCII board, reasoning high 79.5
Grok 4 ASCII board, thinking enabled 79.0
gpt-oss-120b FEN board, reasoning high 78.5
gpt-oss-120b ASCII board, reasoning medium 74.5
o3 FEN board, reasoning high 71.5
Claude Sonnet 4.5 ASCII board, thinking budget 31000 71.5
DeepSeek-V3.2-Exp FEN board, thinking enabled 71.5
GPT-5 mini FEN board, reasoning high 70.0
Claude Sonnet 4 FEN board, thinking budget 31000 69.5
Grok 4 FEN board, thinking enabled 69.0
Gemini 2.5 Pro Jun 2025 ASCII board, thinking budget -1 67.0
Qwen3-235B-A22B-Thinking Jul 2025 ASCII board, reasoning medium, thinking enabled 67.0
Claude Opus 4.1 ASCII board, thinking budget 31000 66.0
Claude Sonnet 4 ASCII board, thinking budget 31000 63.0
DeepSeek-R1 May 2025 FEN board, thinking enabled 61.0
Gemini 2.5 Pro Jun 2025 FEN board, thinking budget -1 59.0
Qwen3-235B-A22B-Thinking Jul 2025 FEN board, reasoning medium, thinking enabled 58.0
GPT-5 mini FEN board, reasoning medium 58.0
DeepSeek-V3.1 FEN board, thinking enabled 58.0
Claude Sonnet 4.5 FEN board, thinking budget 31000 58.0
Claude Opus 4.1 FEN board, thinking budget 31000 56.5
GPT-5 mini ASCII board, reasoning high 54.5
gpt-oss-120b FEN board, reasoning medium 52.0
DeepSeek-V3.1 ASCII board, thinking enabled 51.0
Grok 4 Fast FEN board, thinking enabled 50.5
Kimi K2 Thinking ASCII board, thinking enabled 50.0
Grok 4 Fast ASCII board, thinking enabled 47.5
GPT-5 mini ASCII board, reasoning medium 46.5
DeepSeek-V3.2-Exp ASCII board, thinking enabled 45.0
Kimi K2 Thinking FEN board, thinking enabled 45.0
Qwen3-Max FEN board, thinking non-reasoning model 40.0
DeepSeek-R1 May 2025 ASCII board, thinking enabled 39.5
GLM-4.6 FEN board, thinking enabled 36.5
Qwen3-Max ASCII board, thinking non-reasoning model 36.0
DeepSeek-V3.1 FEN board, thinking disabled 34.0
grok-code-fast-1 ASCII board, thinking enabled 31.0
grok-code-fast-1 FEN board, thinking enabled 31.0
DeepSeek-V3.1 ASCII board, thinking disabled 29.0
Llama 3.3 70B Instruct ASCII board, thinking non-reasoning model 26.0
Kimi K2 Sep 2025 FEN board, thinking non-reasoning model 21.0
GLM-4.5 FEN board, reasoning medium, thinking enabled 17.0
GPT-4o Aug 2024 ASCII board, reasoning non-reasoning model 13.0
Kimi K2 Sep 2025 ASCII board, thinking non-reasoning model 13.0
GLM-4.5 ASCII board, reasoning medium, thinking enabled 11.0
GLM-4.6 ASCII board, thinking enabled 10.5
GPT-4o Aug 2024 FEN board, reasoning non-reasoning model 8.00
Llama 3.3 70B Instruct FEN board, thinking non-reasoning model 8.00

More effort adds a median 10 score

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
GPT-5
FEN board, … ASCII board…
92.0
gpt-oss-120b
FEN board, … ASCII board… ASCII board…
89.0
o3
FEN board, … ASCII board…
79.5
Grok 4
FEN board, … ASCII board…
79.0
Claude Sonnet 4.5
FEN board, … ASCII board…
71.5
DeepSeek-V3.2-Exp
ASCII board… FEN board, …
71.5
GPT-5 mini
ASCII board… FEN board, … FEN board, …
70.0
Claude Sonnet 4
ASCII board… FEN board, …
69.5
Gemini 2.5 Pro
FEN board, … ASCII board…
67.0
Qwen3-235B-A22B-Thinking
FEN board, … ASCII board…
67.0
Claude Opus 4.1
FEN board, … ASCII board…
66.0
DeepSeek-R1
ASCII board… FEN board, …
61.0

Source: Official Game Arena Four-in-a-Row (Generality) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5 ASCII board, reasoning medium 92.0
GPT-5 ASCII board, reasoning high 91.0
gpt-oss-120b ASCII board, reasoning high 89.0
GPT-5 FEN board, reasoning high 88.0
GPT-5 FEN board, reasoning medium 81.0
o3 ASCII board, reasoning high 79.5
Grok 4 ASCII board, thinking enabled 79.0
gpt-oss-120b FEN board, reasoning high 78.5
gpt-oss-120b ASCII board, reasoning medium 74.5
o3 FEN board, reasoning high 71.5
Claude Sonnet 4.5 ASCII board, thinking budget 31000 71.5
DeepSeek-V3.2-Exp FEN board, thinking enabled 71.5
GPT-5 mini FEN board, reasoning high 70.0
Claude Sonnet 4 FEN board, thinking budget 31000 69.5
Grok 4 FEN board, thinking enabled 69.0
Gemini 2.5 Pro Jun 2025 ASCII board, thinking budget -1 67.0
Qwen3-235B-A22B-Thinking Jul 2025 ASCII board, reasoning medium, thinking enabled 67.0
Claude Opus 4.1 ASCII board, thinking budget 31000 66.0
Claude Sonnet 4 ASCII board, thinking budget 31000 63.0
DeepSeek-R1 May 2025 FEN board, thinking enabled 61.0
Gemini 2.5 Pro Jun 2025 FEN board, thinking budget -1 59.0
Qwen3-235B-A22B-Thinking Jul 2025 FEN board, reasoning medium, thinking enabled 58.0
GPT-5 mini FEN board, reasoning medium 58.0
DeepSeek-V3.1 FEN board, thinking enabled 58.0
Claude Sonnet 4.5 FEN board, thinking budget 31000 58.0
Claude Opus 4.1 FEN board, thinking budget 31000 56.5
GPT-5 mini ASCII board, reasoning high 54.5
gpt-oss-120b FEN board, reasoning medium 52.0
DeepSeek-V3.1 ASCII board, thinking enabled 51.0
Grok 4 Fast FEN board, thinking enabled 50.5
Kimi K2 Thinking ASCII board, thinking enabled 50.0
Grok 4 Fast ASCII board, thinking enabled 47.5
GPT-5 mini ASCII board, reasoning medium 46.5
DeepSeek-V3.2-Exp ASCII board, thinking enabled 45.0
Kimi K2 Thinking FEN board, thinking enabled 45.0
Qwen3-Max FEN board, thinking non-reasoning model 40.0
DeepSeek-R1 May 2025 ASCII board, thinking enabled 39.5
GLM-4.6 FEN board, thinking enabled 36.5
Qwen3-Max ASCII board, thinking non-reasoning model 36.0
DeepSeek-V3.1 FEN board, thinking disabled 34.0
grok-code-fast-1 ASCII board, thinking enabled 31.0
grok-code-fast-1 FEN board, thinking enabled 31.0
DeepSeek-V3.1 ASCII board, thinking disabled 29.0
Llama 3.3 70B Instruct ASCII board, thinking non-reasoning model 26.0
Kimi K2 Sep 2025 FEN board, thinking non-reasoning model 21.0
GLM-4.5 FEN board, reasoning medium, thinking enabled 17.0
GPT-4o Aug 2024 ASCII board, reasoning non-reasoning model 13.0
Kimi K2 Sep 2025 ASCII board, thinking non-reasoning model 13.0
GLM-4.5 ASCII board, reasoning medium, thinking enabled 11.0
GLM-4.6 ASCII board, thinking enabled 10.5
GPT-4o Aug 2024 FEN board, reasoning non-reasoning model 8.00
Llama 3.3 70B Instruct FEN board, thinking non-reasoning model 8.00

5 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.

OpenAI Meta
Own leaderboard Not tested since Nov 2025
GPT-5 92.0

Source: Official Game Arena Four-in-a-Row (Generality) leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard
0.00 20.0 40.0 60.0 80.0 100.0 2025 2026 0.00 20.0 40.0 60.0 80.0 100.0 2025 2026

Source: Official Game Arena Four-in-a-Row (Generality) leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Games and long tasks

All 261 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6.2 Astra 100%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress. Llama 4 Scout 7.00
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 12 months ago, in November 2025, for Kimi K2 Thinking. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (23 entries).
  • Pass Models scored 22 models we track, plus 1 other entry on its leaderboard.
  • Warning Newest score The newest came 12 months ago, in November 2025, for Kimi K2 Thinking.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 (#71).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on game-arena.ai: 52 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check