Gemma 3 27B IT

Gemma 3 27B IT is the instruction-tuned flagship of Google's third-generation Gemma family, released on March 12, 2025 as an open-weight successor to Gemma 2. Its weights are published on Hugging Face under the Gemma License, with community-quantized GGUF builds (Q4_K_M, Q5_K_M, Q6_K, Q8_0) and an Ollama package making the model easy to run locally on consumer GPUs. It carries forward the architectural and training advances of Google's Gemini line, positioning the Gemma 3 series as a research-friendly counterpart aimed at developers who want state-of-the-art capability without closed-API constraints.

The model is multimodal on the input side, accepting both text and images while producing text outputs, and supports a context window that extends to roughly 128k tokens. Listed capabilities include structured outputs and function calling, enabling tool-augmented workflows and schema-driven generation alongside chat, reasoning, and math tasks. The combination of a long context, vision-language understanding, and developer-oriented tooling makes Gemma 3 27B IT a practical fit for assistants that need to reason over lengthy documents, mixed media inputs, or agentic pipelines that call external functions.