Gemma 3 1B
Zero-eval
by Google
+
+
+
+
About
Gemma 3 1B was created as an ultra-lightweight open-source model, designed to enable AI capabilities on devices with limited resources. Built with just 1 billion parameters while maintaining instruction-following abilities, it serves edge computing, mobile applications, and scenarios where minimal footprint is essential.
+
+
+
+
Timeline
AnnouncedMar 12, 2025
ReleasedMar 12, 2025
+
+
+
+
Specifications
Training Tokens2.0T
+
+
+
+
License & Family
License
Gemma
Performance Overview
Performance metrics and category breakdown
Overall Performance
18 benchmarks
Average Score
29.9%
Best Score
80.2%
High Performers (80%+)
1+
+
+
+
All Benchmark Results for Gemma 3 1B
Complete list of benchmark scores with detailed information
| IFEval | text | 0.80 | 80.2% | Self-reported | |
| GSM8k | text | 0.63 | 62.8% | Self-reported | |
| Natural2Code | text | 0.56 | 56.0% | Self-reported | |
| MATH | text | 0.48 | 48.0% | Self-reported | |
| HumanEval | text | 0.41 | 41.5% | Self-reported | |
| BIG-Bench Hard | text | 0.39 | 39.1% | Self-reported | |
| FACTS Grounding | text | 0.36 | 36.4% | Self-reported | |
| WMT24++ | text | 0.36 | 35.9% | Self-reported | |
| MBPP | text | 0.35 | 35.2% | Self-reported | |
| Global-MMLU-Lite | text | 0.34 | 34.2% | Self-reported |
Showing 1 to 10 of 18 benchmarks
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+