DialoGPT-large: benchmark results
Microsoft's DialoGPT dialogue-response model for multi-turn conversations, trained on 147 million multi-turn dialogues from Reddit discussion threads. This release retains its distinct size variant; the dialogue count is not a training-token count. Provider: Microsoft. Access: Open.
Aggregate Rating 1117 ± 17, rank #10552 of 10916 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - TruthfulQA MC2 | 50.27 | Accuracy (%) | 77.4 |
| MMLU-by-task - Econometrics | 28.95 | Accuracy (%) | 55.2 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 50.27 | MC2 (%) (0-shot) | 47.8 |
| MMLU-by-task - Machine Learning | 31.25 | Accuracy (%) | 43.7 |
| MMLU-by-task - Astronomy | 33.55 | Accuracy (%) | 36.2 |
| MMLU-by-task - Human Aging | 37.67 | Accuracy (%) | 33.2 |
| MMLU-by-task - TruthfulQA MC1 | 24.97 | Accuracy (%) | 32.8 |
| MMLU-by-task - Conceptual Physics | 32.34 | Accuracy (%) | 31.4 |
| MMLU-by-task - Formal Logic | 25.4 | Accuracy (%) | 28.9 |
| MMLU-by-task - Virology | 31.33 | Accuracy (%) | 28.1 |
| MMLU-by-task - Moral Scenarios | 24.25 | Accuracy (%) | 25.7 |
| MMLU-by-task - High School Chemistry | 27.09 | Accuracy (%) | 25.1 |
Interactive version: theaggregate.ai/model?slug=dialogpt-large · How It Works · Data refreshed daily, snapshot 2026-11-08.