gemma 3 270M by Google

Smallest model of the Gemma 3 family, optimized for on-device use.

Gemma 3 models are well-suited for a variety of text generation understanding tasks, including question answering, summarization, and reasoning.

Supports a context length of 32k tokens.

Q4_0 Quantization Aware Training (QAT) variants preserve similar quality to the full precision model, while significantly reducing the resources required to run it.