1
anthropic/claude-opus-4.1
1.0000
WeirdBench
Benchmark leaderboard
ai-writing-detection
AI Writing Detection
Classify essays from a fixed balanced sample of 50 human-written and 50 AI-generated examples from the AI Generated Essays Dataset. Higher is better.
Higher score is better
2
google/gemini-3-flash-preview
1.0000
3
anthropic/claude-sonnet-4.5
0.9899
4
google/gemma-4-26b-a4b-it
0.9899
5
google/gemma-4-31b-it
0.9899
6
openai/gpt-5.3-chat
0.9899
7
qwen/qwen3.5-122b-a10b
0.9899
8
qwen/qwen3.5-397b-a17b
0.9899
9
openai/gpt-5.1
0.9804
10
x-ai/grok-4.1-fast
0.9804
11
anthropic/claude-opus-4.6
0.9800
12
anthropic/claude-sonnet-4.6
0.9592
13
anthropic/claude-opus-4.5
0.9278
14
google/gemini-3.1-flash-lite-preview
0.8929
15
inception/mercury-2
0.8750
16
openai/gpt-5.4
0.8621
17
anthropic/claude-haiku-4.5
0.8454
18
openai/gpt-oss-120b
0.8421
19
minimax/minimax-m2.5
0.8000
20
meta-llama/llama-4-maverick
0.7321
21
mistralai/mistral-small-2603
0.6849
22
openai/gpt-5.4-mini
0.6803
23
mistralai/mistral-large-2512
0.6667
24
x-ai/grok-4.20-beta
0.6667
25
amazon/nova-micro-v1
0.6438
26
mistralai/mistral-medium-3.1
0.6087
27
deepseek/deepseek-v3.2
0.4819
28
amazon/nova-2-lite-v1
0.4211
29
amazon/nova-pro-v1
0.0357
30
meta-llama/llama-4-scout
0.0241
31
amazon/nova-lite-v1
0.0000
Methodology
How scoring works
Read a fixed deterministic sample of 50 human essays and 50 AI essays from the downloaded Kaggle dataset, prompt the model once per essay to predict whether it is AI-generated, then compute binary classification precision, recall, and F1 for the AI class.
Prompt
Each essay is shown once and the model must return exactly one character: "1" for AI-generated or "0" for human-written, with no explanation.
Score
Higher is better. The benchmark score is the F1 score for detecting AI-generated essays, using label 1 as the positive class.
Execution
Benchmark runners execute locally, read the dataset from disk, use OpenRouter for predictions with reasoning disabled, cache results in Neon by benchmark and model ID, and skip recomputation for models that already have stored scores.