GPT-2 Medium: benchmark results
OpenAI's GPT-2 Medium, the 355M-parameter version of GPT-2, a transformer-based English causal language model, part of the GPT-2/GPT-2-Large/GPT-2-XL family. Provider: OpenAI. Released 2019-05-03. Access: Open.
Aggregate Rating 1146 ± 43, rank #1630 of 1656 rated models, from 110 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| StereoSet | 71.73 | ICAT Score | 80 |
| AI Energy Score (Text Generation) | 5 | Energy Score (1-5) | 79.5 |
| EuroEval Icelandic Common Sense Reasoning | 3.57 | Common Sense Reasoning Average Score (%) | 38.4 |
| Open LLM Leaderboard - MuSR | 38.84 | Score | 34.3 |
| EuroEval Norwegian Knowledge - Idioms NO | 0.39 | MCC (x100) | 26.9 |
| EuroEval Spanish NLU - ScaLA ES | 0.73 | Matthews correlation coefficient (%) | 26.3 |
| EuroEval Norwegian NLU - ScaLA NN | 1.06 | Matthews correlation coefficient (%) | 18.7 |
| Open Portuguese LLM - OAB Exams | 26.2 | Accuracy (%) | 18.5 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Open LLM Leaderboard - GPQA | 26.26 | Score | 16.5 |
| EuroEval Portuguese NLU - ScaLA PT | 0.62 | Matthews correlation coefficient (%) | 16.4 |
| Open LLM Leaderboard - IFEval | 22.08 | Score | 15.1 |
Interactive version: theaggregate.ai/model?slug=gpt-2-medium · How It Works · Data refreshed daily, snapshot 2026-10-20.