Benchmark Almanac

KORGym

o3-mini 0.82 leads since May 2025
Not in model cards

Models play 51 text and visual games designed to test interactive reasoning.

The May 2025 paper compares models across 51 games. The abstract does not state the best score.

Made by
KORGym authors
Released
May 2025
Kind
Puzzles and chess
Level
Everyday
Scored
Settled by events
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Images → Actions
Interaction
Agentic
Environment
KORGym
Refresh
One fixed set
Run by
Anyone
In views
Agents, Vision, Robotics
Strongest model tested
Gemini 2.5 Pro Mar 2025 #104 on the capability index
Last model added
April 2025
Scores refreshed
November 20th, 2026
18
Entries on its own leaderboard, 12 models we track

o3-mini leads with 0.82 among models released by April 2025 o3-mini leads with 0.82 among models released by April 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. 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 April 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 o3-mini Thinking 0.82
  2. 2 Gemini 2.5 Pro Mar 2025, Thinking 0.79
  3. 3 o1 Thinking 0.73
  4. 4 Doubao-1-5-thinking-pro Thinking 0.72
  5. 5 DeepSeek-R1 (snapshot unspecified) Thinking 0.71
  6. 6 Claude 3.7 Sonnet Thinking 0.62
  7. 7 Qwen3-32B Thinking 0.60
  8. 8 DeepSeek-V3 Mar 2025, Non-thinking 0.42
  9. 9 DeepSeek-R1-Distill-Qwen-32B Thinking 0.39
  10. 10 Gemini 2.0 Flash Thinking Jan 2025, Thinking 0.38
  11. 11 Qwen-QwQ Thinking 0.28
  12. 12 Gemini 2.0 Flash Feb 2025, Non-thinking 0.26
  13. 13 GPT-4o (snapshot unspecified) Non-thinking 0.22
  14. 14 Doubao-1.5-pro Non-thinking 0.20
  15. 15 Qwen2.5-72B Non-thinking 0.18
  16. 16 Qwen2.5-32B-Instruct Non-thinking 0.16
  17. 17 DeepSeek-R1-Distill-Qwen-7B Non-thinking 0.11
  18. 18 Qwen2.5-7B-Instruct Non-thinking 0.08

Source: Official KORGym leaderboard

#EntryScoreSettingDateSource
1 o3-mini 0.82 Thinking Jan 2025 Official leaderboard
2 Gemini 2.5 Pro Mar 2025 0.79 Thinking Mar 2025 Official leaderboard
3 o1 0.73 Thinking Dec 2024 Official leaderboard
4 Doubao-1-5-thinking-pro 0.72 Thinking Official leaderboard
5 DeepSeek-R1 (snapshot unspecified) 0.71 Thinking Official leaderboard
6 Claude 3.7 Sonnet 0.62 Thinking Feb 2025 Official leaderboard
7 Qwen3-32B 0.60 Thinking Apr 2025 Official leaderboard
8 DeepSeek-V3 Mar 2025 0.42 Non-thinking Mar 2025 Official leaderboard
9 DeepSeek-R1-Distill-Qwen-32B 0.39 Thinking Jan 2025 Official leaderboard
10 Gemini 2.0 Flash Thinking Jan 2025 0.38 Thinking Jan 2025 Official leaderboard
11 Qwen-QwQ 0.28 Thinking Official leaderboard
12 Gemini 2.0 Flash Feb 2025 0.26 Non-thinking Feb 2025 Official leaderboard
13 GPT-4o (snapshot unspecified) 0.22 Non-thinking Official leaderboard
14 Doubao-1.5-pro 0.20 Non-thinking Jan 2025 Official leaderboard
15 Qwen2.5-72B 0.18 Non-thinking Sep 2024 Official leaderboard
16 Qwen2.5-32B-Instruct 0.16 Non-thinking Official leaderboard
17 DeepSeek-R1-Distill-Qwen-7B 0.11 Non-thinking Jan 2025 Official leaderboard
18 Qwen2.5-7B-Instruct 0.08 Non-thinking Official leaderboard
#ModelScoreSettingDateSource
1 o3-mini 0.82 Thinking Jan 2025 Official leaderboard
2 Gemini 2.5 Pro Mar 2025 0.79 Thinking Mar 2025 Official leaderboard
3 o1 0.73 Thinking Dec 2024 Official leaderboard
4 Claude 3.7 Sonnet 0.62 Thinking Feb 2025 Official leaderboard
5 Qwen3-32B 0.60 Thinking Apr 2025 Official leaderboard
6 DeepSeek-V3 Mar 2025 0.42 Non-thinking Mar 2025 Official leaderboard
7 DeepSeek-R1-Distill-Qwen-32B 0.39 Thinking Jan 2025 Official leaderboard
8 Gemini 2.0 Flash Thinking Jan 2025 0.38 Thinking Jan 2025 Official leaderboard
9 Gemini 2.0 Flash Feb 2025 0.26 Non-thinking Feb 2025 Official leaderboard
10 Doubao-1.5-pro 0.20 Non-thinking Jan 2025 Official leaderboard
11 Qwen2.5-72B 0.18 Non-thinking Sep 2024 Official leaderboard
12 DeepSeek-R1-Distill-Qwen-7B 0.11 Non-thinking Jan 2025 Official leaderboard

Rankings by part

We read 47 parts from the benchmark’s own leaderboard in 5 groups: strategy, spatial, puzzle, math and control. Each ranks models on one part of the test. Gemini 2.5 Pro leads 15 of them, and 5 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Control and Interaction

12 models we track have a score here.

  1. 1 o3-mini 0.77
  2. 2 o1 ↑1 0.74
  3. 3 Gemini 2.5 Pro Mar 2025 ↓1 0.63
  4. 4 Qwen3-32B ↑1 0.58
  5. 5 Claude 3.7 Sonnet ↓1 0.50
  6. 6 DeepSeek-R1-Distill-Qwen-32B ↑1 0.45
  7. 7 DeepSeek-V3 Mar 2025 ↓1 0.35
  8. 8 Gemini 2.0 Flash Thinking Jan 2025 0.25
  9. 9 Gemini 2.0 Flash Feb 2025 0.24
  10. 10 Qwen2.5-72B ↑1 0.18

Number one over time

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

OpenAI
Own leaderboard Not tested since May 2025
o3-mini 0.82

Source: Official KORGym 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 0.20 0.40 0.60 0.80 1.00 2025 0.00 0.20 0.40 0.60 0.80 1.00 2025

Source: Official KORGym 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 19 months ago, in April 2025, for Qwen3-32B. 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 (18 entries).
  • Pass Models scored 12 models we track, plus 6 other entries on its leaderboard.
  • Warning Newest score The newest came 19 months ago, in April 2025, for Qwen3-32B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 2.5 Pro (#104).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on razor233.github.io: 13 rows for models we track and 6 other entries 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