xglm-2.9B: benchmark results
Meta FAIR's XGLM 2.9B, a dense multilingual decoder-only base model trained on 500 billion sub-tokens across 30 languages; this is not an instruction-tuned release. Provider: Other. Access: Open.
Aggregate Rating 1209 ± 48, rank #9560 of 10916 rated models, from 11 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Greek MMLU - Social Sciences | 34.94 | Accuracy (%) | 19.8 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Greek MMLU - Greek-Specific | 32.9 | Accuracy (%) | 16 |
| Open Portuguese LLM - BLUEX | 19.75 | Accuracy (%) | 8.9 |
| Open Portuguese LLM - ENEM | 19.24 | Accuracy (%) | 8.4 |
| Open Portuguese LLM - OAB Exams | 22.69 | Accuracy (%) | 6.8 |
| Greek MMLU | 32.68 | Accuracy (%) | 6.2 |
| Open Portuguese LLM - ASSIN2 RTE | 33.33 | Macro F1 (%) | 4.9 |
| Greek MMLU - STEM | 33.25 | Accuracy (%) | 3.7 |
| Greek MMLU - Other | 26.92 | Accuracy (%) | 1.2 |
| Greek MMLU - Humanities | 29.39 | Accuracy (%) | 0 |
Interactive version: theaggregate.ai/model?slug=xglm-2-9b · How It Works · Data refreshed daily, snapshot 2026-11-08.