T5 (11B): benchmark results
Google's T5-11B, the 11-billion-parameter checkpoint of the Text-To-Text Transfer Transformer, reframing NLP tasks such as translation, summarization and classification into a unified text-to-text format. Provider: Google. Released 2019-10-23. Access: Open.
Aggregate Rating 1054 ± 32, rank #1631 of 1635 rated models, from 34 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BoolQ | 91.2 | Accuracy (%) | 100 |
| Epoch AI - Superglue | 88.9 | Score | 100 |
| GLUE SST-2 | 97.5 | Accuracy | 99.4 |
| GLUE MNLI (Mismatched) | 91.7 | Mismatched | 96.6 |
| RealToxicityPrompts | 0.01 | Toxic fraction | 95.1 |
| GLUE QNLI | 96.7 | Accuracy | 89.1 |
| Epoch AI - Common Sense Qa 2 | 67.8 | Score | 83.3 |
| GLUE CoLA | 70.8 | Accuracy | 73.2 |
| HELM Classic - LegalSupport | 55.83 | Exact Match (%) | 63.2 |
| HELM Classic - BoolQ | 76.1 | Exact Match (%) | 57.6 |
| HELM Classic - Synthetic Reasoning Abstract | 19.59 | Exact Match (%) | 39.7 |
| HELM Classic - MMLU | 29.03 | Exact Match (%) | 36.4 |
Interactive version: theaggregate.ai/model?slug=t5-11b · How It Works · Data refreshed daily, snapshot 2026-11-05.