BitNet b1.58 2B4T.
Microsoft Apr 2025 Open weights Small
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BoolQ | 80.2% | #16 of 28 | Lab model card | |
| WinoGrande | 71.9% | #37 of 47 | Lab model card | |
| CommonSenseQA | 71.6% | #15 of 19 | Lab model card | |
| TruthfulQA | 45.3% | #24 of 30 | Lab model card | |
| ARC-Challenge | 49.9% | #44 of 49 | Lab model card | |
| ARC-Easy | 74.8% | #21 of 23 | Lab model card | |
| PIQA | 77.1% | #35 of 37 | Lab model card | |
| MMLU | 53.2% | #124 of 131 | Lab model card | |
| HellaSwag | 68.4% | #51 of 53 | Lab model card | |
| IFEval | 53.5% | #133 of 135 | Lab model card | |
| TriviaQA | 33.6% | #40 of 40 | Lab model card | |
| OpenBookQA | 41.6% | #24 of 24 | Lab model card | |
| MT-Bench | 5.85 | #23 of 23 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HumanEval+ | 38.4% | #27 of 27 | Lab model card |
The Self-Reported Index fits one score to the 15 capability results Microsoft reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 92.7 to 99.1. How we compute it.
The benchmarks the lab chose to headline, with the scores it gave.
Every result in the longer documents the lab published with it.