MobileLLM-R1-950M: benchmark results
Meta MobileLLM-R1 950M is the final reasoning SFT checkpoint for mathematics, programming and scientific problems, separate from its Base release. The publisher specifies 22 layers, 24 attention heads, 6 key/value heads and shared input/output embeddings. Provider: Meta. Released 2025-09-12. Access: Open.
Aggregate Rating 1191 ± 36, rank #9663 of 10916 rated models, from 142 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Icelandic Common Sense Reasoning | 1.51 | Common Sense Reasoning Average Score (%) | 29.1 |
| EuroEval Spanish NLU - ScaLA ES | 0.83 | Matthews correlation coefficient (%) | 27 |
| EuroEval Lithuanian NLU - ScaLA LT | 1.63 | Matthews correlation coefficient (%) | 25.3 |
| EuroEval Faroese NLU - ScaLA FO | 0.27 | Matthews correlation coefficient (%) | 21.7 |
| EuroEval Lithuanian Common Sense Reasoning | 0.1 | Common Sense Reasoning Average Score (%) | 20.9 |
| EuroEval Latvian NLU - ScaLA LV | 0.02 | Matthews correlation coefficient (%) | 20 |
| EuroEval Portuguese NLU - ScaLA PT | 1.05 | Matthews correlation coefficient (%) | 19.8 |
| EuroEval Polish Common Sense Reasoning | 0.42 | Common Sense Reasoning Average Score (%) | 19.3 |
| EuroEval Estonian Common Sense Reasoning | 0.32 | Common Sense Reasoning Average Score (%) | 18.4 |
| EuroEval Portuguese Common Sense Reasoning | 1.21 | Common Sense Reasoning Average Score (%) | 18.3 |
| EuroEval Dutch Common Sense Reasoning | 1.09 | Common Sense Reasoning Average Score (%) | 16 |
| EuroEval Norwegian NLU - ScaLA NB | 0.87 | Matthews correlation coefficient (%) | 15.9 |
Interactive version: theaggregate.ai/model?slug=mobilellm-r1-950m · How It Works · Data refreshed daily, snapshot 2026-11-08.