opt-iml-max-30B: benchmark results
Meta OPT-IML-Max is the instruction-tuned OPT release trained on approximately 2,000 tasks from eight benchmark collections. The Max training mix remains distinct from OPT-IML, which reserves tasks for evaluation, and from the pretrained OPT base model. Provider: Meta. Access: Open.
Aggregate Rating 1245 ± 16, rank #9288 of 10916 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - College Mathematics | 36 | Accuracy (%) | 80.2 |
| MMLU-by-task - College Physics | 27.45 | Accuracy (%) | 63 |
| MMLU-by-task - High School Physics | 31.79 | Accuracy (%) | 61.3 |
| MMLU-by-task - College Computer Science | 42 | Accuracy (%) | 61.2 |
| MMLU-by-task - High School World History | 70.04 | Accuracy (%) | 60.6 |
| MMLU-by-task - Moral Scenarios | 27.15 | Accuracy (%) | 58.3 |
| MMLU-by-task - Astronomy | 46.71 | Accuracy (%) | 55.5 |
| MMLU-by-task - High School Macroeconomics | 44.62 | Accuracy (%) | 54.9 |
| MMLU-by-task - Formal Logic | 30.95 | Accuracy (%) | 52.2 |
| MMLU-by-task - Virology | 40.96 | Accuracy (%) | 51.1 |
| MMLU-by-task - Elementary Mathematics | 28.84 | Accuracy (%) | 50.7 |
| MMLU-by-task - Moral Disputes | 49.42 | Accuracy (%) | 48.7 |
Interactive version: theaggregate.ai/model?slug=opt-iml-max-30b · How It Works · Data refreshed daily, snapshot 2026-11-08.