Skip to content

Benchmark

ZEROBench

Scores are percentages. Higher is better.

Models measured
84
Scores recorded
90
Sources
3
IQ weight
2.09

Score distribution

5.4%
Mean
2%
Median
7.8%
Std dev
0%
10th percentile
13.4%
90th percentile
0%
Min
0%43.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.09 for chat models, measured over 84 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Grok 4· ZeroBench1.4%
52Claude Opus 4· ZeroBench1.2%
53Claude Sonnet 4.5· ZeroBench1.2%
54GPT-5 nanohigh reasoning· ZeroBench1%
55Qwen3.5 Omni FlashQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.1%
56GPT 4.1· ZeroBench0.6%
57DeepSeek V4.1 Flash 1017· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.51%
58Muse Spark Open· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.5%
59DeepSeek V4.1 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.49%
60Hy4 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.458%
61Kimi K3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.41%
62Claude Sonnet 3.7· ZeroBench0.4%
63Llama 4 Maverick· ZeroBench0.4%
64Kimi K3.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.39%
65DeepSeek V4 Flash Vision Exp· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.35%
66Muse Spark· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.33%
67Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.18%
68Inkling 2 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.17%
69Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.11%
70Kimi K2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.11%
71Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.1%
72Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.09%
73Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.08%
74Seed 2.0 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim0.07%
75Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.0.04%
76Claude 3 Haiku· ZeroBench0%
77Claude 3 Opus· ZeroBench0%
78Claude 3 Sonnet· ZeroBench0%
79Claude 3.5 Sonnet· ZeroBench0%
80Claude 3.5 Sonnet v2· ZeroBench0%
81Gemini 1.5 Flash· ZeroBench0%
82Gemini 1.5 Pro· ZeroBench0%
83GPT 4o· ZeroBench0%
84GPT-4o mini· ZeroBench0%
85llama-3.2-90b· ZeroBench0%
86o1· ZeroBench0%
87O1 Pro· ZeroBench0%
88Pixtral Large· ZeroBench0%
89Qwen2-VL (72B) Instruct· ZeroBench0%
90Reka Edge· ZeroBench0%

Showing 51–90 of 90 results.

Data credits: Data by LLM Stats · 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.