Gemini Nano 1
Google DeepMind · 2023-12-06
Gemini Nano 1 is the smaller of Google's two original Nano-tier models for on-device deployment, targeting lower-memory devices. Google reported a rounded 1.8B-parameter size, said the Nano series was distilled from larger Gemini models and 4-bit quantized for deployment, and introduced it on December 6, 2023. The Gemini 1.0 report documents a 32,768-token context and evaluates Nano 1 on both language and pixel-input tasks.
Benchmark scores
| Benchmark | Score |
|---|---|
| BBH (BIG-Bench Hard) | 34.8 |
| BoolQ | 71.6 |
| Capability | 85.5 |
| MATH | 13.5 |
| MathVista (mini) | 27.3 |
| MMLU | 45.9 |
| MMMU | 26.3 |