Skip to content

Benchmark

NYT Connections – Extended Version (Lech Mazur)

Scores are percentages.

Models measured
106
Scores recorded
138
Sources
1
IQ weight
2.44

Score distribution

63.5%
Mean
75.1%
Median
30.5%
Std dev
14.5%
10th percentile
94.4%
90th percentile
4.1%
Min
4.1%99.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.44 for chat models, measured over 106 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
101Hy3 (Deep Infra)high· Lech Mazur suite data41.2%
102Step 3.7 Flashhigh reasoning· Lech Mazur suite data39.7%
103Claude Opus 4.7high reasoning· Lech Mazur suite data39%
104Claude Sonnet 4.5thinking, 16k· Lech Mazur suite data37.3%
105DeepSeek V3.2· Lech Mazur suite data36.7%
106Claude Sonnet 4.5no reasoning· Lech Mazur suite data35.8%
107MiMo V2.5 Pro· Lech Mazur suite data34.4%
108Qwen3 Max· Lech Mazur suite data30.1%
109Step 3.5 Flash· Lech Mazur suite data28.4%
110MiniMax M3.1 Flashhigh reasoning· Lech Mazur suite data28.3%
111Mistral Large 4high· Lech Mazur suite data27.4%
112MiMo V2 Pro· Lech Mazur suite data25.8%
113MiniMax M2.7· Lech Mazur suite data24.7%
114ERNIE 5.1· Lech Mazur suite data23.4%
115GPT 5.5no reasoning· Lech Mazur suite data22%
116GPT 5.4no reasoning· Lech Mazur suite data17.8%
117LongCat-Flash-Thinking· Lech Mazur suite data17.7%
118Hy3 (Deep Infra)· Lech Mazur suite data17.2%
119MiniMax M2.5· Lech Mazur suite data16.8%
120Trinity Large Thinking· Lech Mazur suite data16.5%
121Gemma 4 31B IT· Lech Mazur suite data15.7%
122Nemotron 3 Super· Lech Mazur suite data15.4%
123MiniMax M2· Lech Mazur suite data14.8%
124GPT 5.2no reasoning· Lech Mazur suite data14.5%
125Claude Haiku 4.5· Lech Mazur suite data14.3%
126Mistral Medium 3.5high· Lech Mazur suite data12.9%
127Grok 4.1 Fastnon-reasoning· Lech Mazur suite data12.3%
128MiniMax M2.1· Lech Mazur suite data11.2%
129Claude Opus 4.7no reasoning· Lech Mazur suite data10.8%
130ERNIE 5.0· Lech Mazur suite data10.3%
131Grok 4.20Non-Reasoning· Lech Mazur suite data8.6%
132DeepSeek V3.2no reasoning· Lech Mazur suite data8.2%
133Gemini 3.1 Flash Lite· Lech Mazur suite data8.2%
134Llama 4 Maverick· Lech Mazur suite data8%
135Mistral Large 3· Lech Mazur suite data7.5%
136Hy3 (Deep Infra)non-thinking· Lech Mazur suite data6.9%
137Mistral Medium 3.1· Lech Mazur suite data6.5%
138ling-2.6-1t· Lech Mazur suite data4.1%

Showing 101–138 of 138 results.