Skip to content

Benchmark

RecreationBench paper · Ubuntu · Visual assertion score

Scores are percentages. Higher is better.

Models and systems measured
10
Scores recorded
10
Sources
1
IQ weight
0.99

Score distribution

23.6%
Mean
26.6%
Median
14.2%
Std dev
6.9%
10th percentile
35.9%
90th percentile
2.2%
Min
2.2%52.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.99 for chat models, measured over 8 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1GPT-6 AstraRecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · max reasoning effort per site data · 272k context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs52.6%
2GPT-5.6 SolRecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · max reasoning effort per site data · 272k context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs35.9%
3Claude Opus 5RecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · max reasoning effort per site data · 1m context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs30.8%
4Kimi K3RecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · max reasoning effort per site data · 1m context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs23.4%
5Claude Opus 4.8RecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · max reasoning effort per site data · 1m context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs18.2%
6GLM 5.3RecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · max reasoning effort per site data · 1m context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs10.5%
7Gemini 3.7 FlashRecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · high reasoning effort per site data · 1m context · 64k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs6.9%
8Qwen3.7 PlusRecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · thinking enabled per site data · 256k context · 32k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.2.2%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Grok 4.6RecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · xhigh reasoning effort per site data · 500k context · 128k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs26.6%
Qwen3.8 Max 0902RecreationWorld benchmark-owner paper run · 250 held-out tasks · GUI exploration + coding tools + visual verification · xhigh reasoning effort per site data · 256k context · 32k max output tokens · Ubuntu · nominal 50-task split · platform-specific score· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.28.7%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.