Skip to content

Benchmark

RecreationBench paper · Overall · Programmatic assertion score

Scores are percentages. Higher is better.

Models and systems measured
10
Scores recorded
10
Sources
1
IQ weight
1.48

Score distribution

34.4%
Mean
35.5%
Median
12.5%
Std dev
24.9%
10th percentile
46%
90th percentile
9.2%
Min
9.2%58.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.48 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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs58.2%
2Claude 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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs46%
3GPT-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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs40.6%
4Claude 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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs32.4%
5Kimi 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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs32.1%
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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs26.3%
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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs24.9%
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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.9.2%

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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs39%
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 · Overall · 250 held-out tasks · equal-weight platform macro average· Reported by QwenLM / RecreationWorld authors· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.35.5%

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.