GPT-2: benchmark results
OpenAI's GPT-2, the 124M-parameter checkpoint (Hugging Face gpt2) and the smallest of the four GPT-2 sizes, a transformer-based English causal language model; GPT-2 Medium, Large and XL have their own rows. Provider: OpenAI. Released 2019-02-14. Access: Open.
Aggregate Rating 1098 ± 32, rank #1647 of 1656 rated models, from 126 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| StereoSet | 72.97 | ICAT Score | 100 |
| Open LLM Leaderboard - MuSR | 44.71 | Score | 81.9 |
| AI Energy Score (Text Generation) | 5 | Energy Score (1-5) | 79.5 |
| EuroEval Faroese NLU - ScaLA FO | 0.79 | Matthews correlation coefficient (%) | 31.1 |
| Open Portuguese LLM - FaQuAD NLI | 45.18 | Macro F1 (%) | 27.5 |
| EuroEval Spanish NLU - ScaLA ES | 0.01 | Matthews correlation coefficient (%) | 19.3 |
| EuroEval Portuguese NLU - ScaLA PT | 0.91 | Matthews correlation coefficient (%) | 18.2 |
| EuroEval Norwegian NLU - ScaLA NN | 0.84 | Matthews correlation coefficient (%) | 16.3 |
| EuroEval Italian NLU - ScaLA IT | 1.07 | Matthews correlation coefficient (%) | 14.9 |
| EuroEval Icelandic Common Sense Reasoning | -0.05 | Common Sense Reasoning Average Score (%) | 13.3 |
| Open LLM Leaderboard - GPQA | 26.01 | Score | 13.2 |
| EuroEval Danish Knowledge - Danske Talemaader | 0.55 | MCC (x100) | 12.9 |
Interactive version: theaggregate.ai/model?slug=gpt-2 · How It Works · Data refreshed daily, snapshot 2026-10-20.