We’re excited to announce that Google’s Gemma 3 fashions are coming to Databricks, beginning with the Gemma 3 12B mannequin, which is now natively out there throughout all clouds. Textual content capabilities can be found as we speak, with multimodal assist rolling out quickly.
Gemma 3 12B is optimized for enterprise workloads, hanging the best steadiness between functionality and computational effectivity. It excels at core use circumstances like doc processing, content material evaluation, code era, and conversational AI, making it a robust match for production-grade functions.
Databricks has lengthy been the platform the place enterprises handle and analyze unstructured information at scale. As enterprises join that information with giant language fashions to construct AI brokers, the necessity for environment friendly, high-quality fashions with an inexpensive worth level has grown quickly. Gemma 3 12B fills a important hole—providing open, high-quality multimodal capabilities that energy doc AI and visible query answering use circumstances. Mixed with Databricks’ unified platform for unstructured information and mannequin improvement, groups can construct and deploy production-grade AI sooner and extra affordably.
Enhanced high quality and multimodal capabilities
Gemma 3 12B offers a gorgeous steadiness of dimension and high quality:
- Broad capabilities: Gemma 3 12B has a big 128K context window that helps lengthy paperwork as inputs. It additionally helps over 140 languages, making it superb for international companies.
- Top quality for advanced text-based duties: Gemma 3 12B excels in language understanding and mathematical problem-solving. This makes it well-suited for enterprise functions that require subtle textual content evaluation and logical reasoning.
- Empowers doc AI and visible query answering use circumstances: The extra picture modality unlocks use circumstances that require enter from unstructured information, resembling extracting data from tables in paperwork, receipt images, and extra.
Get began with Gemma 3 12B on Databricks
To get a way of whether or not Gemma 3 12B would fit your use case, attempt it out in AI Playground.
You’ll be able to question the mannequin serving endpoint as properly. For each AI Playground and the mannequin serving endpoint, multimodal capabilities are coming quickly.
Moreover, the newly launched MLflow 3 permits you to consider the mannequin extra comprehensively throughout your particular datasets.
You may also run scalable batch inference by sending a SQL question to your desk.