opt-30B: benchmark results
Meta's OPT-30B, part of the Open Pretrained Transformer suite of decoder-only models from 125M to 175B parameters, designed to match GPT-3-class model sizes and performance. Provider: Meta. Released 2022-05-03. Access: Open.
Aggregate Rating 1166 ± 13, rank #1610 of 1656 rated models, from 90 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - High School Statistics | 42.13 | Accuracy (%) | 62.8 |
| MMLU-by-task - Moral Scenarios | 27.6 | Accuracy (%) | 62 |
| ToolBench - WebShop Long | 0 | Task Score | 44.2 |
| MMLU-by-task - College Computer Science | 35 | Accuracy (%) | 42 |
| MMLU-by-task - HellaSwag | 54.92 | Accuracy (%) | 40 |
| MMLU-by-task - High School Physics | 28.48 | Accuracy (%) | 39.1 |
| MMLU-by-task - High School US History | 30.88 | Accuracy (%) | 34.6 |
| MMLU-by-task - Security Studies | 36.33 | Accuracy (%) | 34 |
| MMLU-by-task - World Religions | 35.67 | Accuracy (%) | 33.8 |
| MMLU-by-task - ARC Challenge | 39.42 | Accuracy (%) | 33.6 |
| MMLU-by-task - College Mathematics | 29 | Accuracy (%) | 32.7 |
| ToolBench - Google Sheets | 8.6 | Task Score | 32.6 |
Interactive version: theaggregate.ai/model?slug=opt-30b · How It Works · Data refreshed daily, snapshot 2026-10-20.