PuzzleWorld
Models solve 667 puzzle-hunt problems that require varied reasoning strategies.
The June 2025 paper compares language models on 667 puzzle-hunt problems. The abstract does not state the best score.
- Made by
- PuzzleWorld authors
- Released
- June 2025
- Kind
- Puzzles and chess
- Level
- Competition
- Scored
- Settled by events
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Exams and contests
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Strongest model tested
-
Gemini 3 Pro #56 on the capability index
- Last model added
- November 2025
- Scores refreshed
- November 20th, 2026
Gemini 3 Pro leads with 18% among models released by November 2025 Gemini 3 Pro leads with 18% among models released by November 2025
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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 November 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1
Gemini 3 Pro 18%
- 2 o3 14.2%
- 3 Gemini 2.5 Pro (snapshot unspecified) 7.65%
- 4 Claude Opus 4 4.50%
- 5 Grok 4 3.33%
- 6 GPT-4o (snapshot unspecified) 1.83%
- 7
QVQ-72B-Preview 1.36%
- 8 Kimi VL A3B 1.33%
- 9
InternVL3-78B 0.89%
Source: The PuzzleWorld paper
- 1
Gemini 3 Pro 18%
- 2 o3 14.2%
- 3 Claude Opus 4 4.50%
- 4 Grok 4 3.33%
- 5
QVQ-72B-Preview 1.36%
- 6
InternVL3-78B 0.89%
Source: The PuzzleWorld paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 18% | Nov 2025 | Official leaderboard | |
| 2 | o3 | 14.2% | Apr 2025 | Official leaderboard | |
| 3 | Gemini 2.5 Pro (snapshot unspecified) closed weights | 7.65% | Official leaderboard | ||
| 4 | Claude Opus 4 | 4.50% | May 2025 | Official leaderboard | |
| 5 | Grok 4 | 3.33% | Jul 2025 | Official leaderboard | |
| 6 | GPT-4o (snapshot unspecified) closed weights | 1.83% | Official leaderboard | ||
| 7 | | 1.36% | Dec 2024 | Official leaderboard | |
| 8 | Kimi VL A3B open weights | 1.33% | Official leaderboard | ||
| 9 | | 0.89% | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 18% | Nov 2025 | Official leaderboard | |
| 2 | o3 | 14.2% | Apr 2025 | Official leaderboard | |
| 3 | Claude Opus 4 | 4.50% | May 2025 | Official leaderboard | |
| 4 | Grok 4 | 3.33% | Jul 2025 | Official leaderboard | |
| 5 | | 1.36% | Dec 2024 | Official leaderboard | |
| 6 | | 0.89% | Apr 2025 | Official leaderboard |
Rankings by part
We read 6 parts from the benchmark’s paper, each ranking models on one part of the test, such as Text answer accuracy, Text stepwise accuracy and Visual answer accuracy. Gemini 3 Pro leads 4 of them, and o3 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.
Text answer accuracy
6 models we track have a score here.
- 1
Gemini 3 Pro 20.3%
- 2 o3 15.2%
- 3 Claude Opus 4 4.20%
- 4 Grok 4 3.85%
- 5
QVQ-72B-Preview 1.33%
- 6
InternVL3-78B 0.83%
Text stepwise accuracy
6 models we track have a score here.
- 1 o3 ↑1 39.9%
- 2
Gemini 3 Pro ↓1 39.3%
- 3
QVQ-72B-Preview ↑2 29.3%
- 4 Claude Opus 4 ↓1 23.8%
- 5
InternVL3-78B ↑1 14.8%
- 6 Grok 4 ↓2 13.6%
Visual answer accuracy
6 models we track have a score here.
- 1
Gemini 3 Pro 14.7%
- 2 o3 8.96%
- 3 Claude Opus 4 4.04%
- 4 Grok 4 3.70%
- 5
QVQ-72B-Preview 0.63%
- 6
InternVL3-78B 0.47%
Visual stepwise accuracy
6 models we track have a score here.
- 1
Gemini 3 Pro 38.8%
- 2 o3 33.4%
- 3
QVQ-72B-Preview ↑2 28%
- 4 Claude Opus 4 ↓1 22.6%
- 5
InternVL3-78B ↑1 14.5%
- 6 Grok 4 ↓2 14.2%
Structured answer accuracy
6 models we track have a score here.
- 1
Gemini 3 Pro 20.3%
- 2 o3 13.5%
- 3 Claude Opus 4 4.37%
- 4 Grok 4 1.56%
- 5
QVQ-72B-Preview 1.18%
- 6
InternVL3-78B 1.15%
Structured stepwise accuracy
6 models we track have a score here.
- 1 o3 ↑1 41.3%
- 2
Gemini 3 Pro ↓1 40%
- 3
QVQ-72B-Preview ↑2 32.4%
- 4 Claude Opus 4 ↓1 26.9%
- 5
InternVL3-78B ↑1 18%
- 6 Grok 4 ↓2 11.2%
2 models have held number one in the benchmark’s paper since its release in June 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The PuzzleWorld paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official PuzzleWorld 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
- 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
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 12 months ago, in November 2025, for Gemini 3 Pro. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (9 entries).
- Pass Models scored 6 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 12 months ago, in November 2025, for Gemini 3 Pro.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
- Pass Headroom The best is 18%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 6 rows for models we track and 3 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.