gemma-1.1-2b-it

Gemma 1.1 2B IT is Google's instruction-tuned update to the original Gemma 2B model, released in April 2024 as part of the Gemma open-weight family built on the same research foundation as Gemini. It is a decoder-only Transformer language model with 2 billion parameters, trained on 6 trillion tokens spanning web text, code, and mathematics, and further aligned using a novel RLHF method that improves quality, coding ability, factuality, instruction following, and multi-turn conversation.

The model supports an 8192-token context window and uses a chat template with role-based turn delimiters for conversational use. Gemma 1.1 also fixes a multi-turn conversation bug from the prior release and reduces repetitive "Sure," response prefixes. Weights are available under the Gemma license on Hugging Face in bfloat16, with optional 4-bit/8-bit quantization for lower-memory deployment.

At 2B parameters, Gemma 1.1 2B IT is designed for resource-constrained environments such as laptops, desktops, and modest cloud instances. It suits question answering, summarization, lightweight coding assistance, and general chat applications where a small footprint matters more than frontier-scale reasoning performance.