Gemma 2

Gemma 2 is a compact, production-ready successor to the original Gemma models. Offered in multiple sizes—most commonly 9B and 27B—with base and instruction-tuned variants, it delivers high quality at low latency and cost. The models support long-context prompts, function/tool calling, and structured (JSON) outputs, making them reliable in RAG pipelines and agent workflows. They’re optimized for modern GPUs/TPUs, with strong throughput per dollar, and run well under common inference stacks with 8/4-bit quantization. Typical uses include chat and copilots, document QA and summarization, analytics over long reports, and coding assistants. Gemma 2 ships as open weights under Google’s Gemma license and is available for both local deployment and managed hosting.