bloomz-1b7: benchmark results
BigScience's BLOOMZ-1b7 multilingual instruction-tuned checkpoint, distinct from the larger BLOOMZ-7b1 release. Provider: BigScience. Access: Open.
Aggregate Rating 1289 ± 29, rank #8868 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Greek MMLU - Other | 35.94 | Accuracy (%) | 15.4 |
| Greek MMLU - STEM | 35.66 | Accuracy (%) | 14.8 |
| UrduMMLU (English Prompt) | 29.48 | Overall exact-match accuracy (%) on UrduMMLU's Urdu multiple | 13.8 |
| Greek MMLU | 34.66 | Accuracy (%) | 11.1 |
| UrduMMLU (Urdu Prompt) - Social Sciences | 28.91 | Social Sciences (7,959 questions in 11 subjects, including P | 10.7 |
| Greek MMLU - Greek-Specific | 31.53 | Accuracy (%) | 8.6 |
| Greek MMLU - Social Sciences | 33.15 | Accuracy (%) | 8.6 |
| Greek MMLU - Humanities | 34.19 | Accuracy (%) | 8 |
| UrduMMLU (Urdu Prompt) - General Knowledge | 25.92 | General knowledge (1,362 questions, the Other domain) exact- | 7.1 |
| UrduMMLU (Urdu Prompt) - Profession | 27.34 | Profession (973 questions: professional development and stud | 7.1 |
| UrduMMLU (Urdu Prompt) | 23.04 | Overall exact-match accuracy (%) on UrduMMLU's Urdu multiple | 3.6 |
| UrduMMLU (Urdu Prompt) - Humanities | 20.78 | Humanities (10,982 questions: ethics, fine arts, Islamic stu | 3.6 |
Interactive version: theaggregate.ai/model?slug=bloomz-1b7 · How It Works · Data refreshed daily, snapshot 2026-11-08.