Databricks

Can object storage sit underneath a transactional database and make it easier for agents to work with? This question is what started Lakebase Postgres. The answer does not just depend on how fast your object store is, but rather where you place the source of truth. See how we architected Lakebase Postgres and why when Postgres is architected this way, it becomes an evolution of traditional OLTP systems that’s built to handle agentic workloads: https://lnkd.in/gJS6kPeM

“Genie is a game changer. Genie helps me go from ideation to production faster, quicker, and without any errors.” Hear directly from users on how they’re using Genie to ask better questions, get to answers faster, and work across the tools they already use 🧞 https://lnkd.in/gjKVKVEC

Databricks has been named to Madrona's 2026 Intelligent Applications 40 list! We're honored to be recognized among the top private companies building with and enabling the AI applications shaping our future. Databricks is also the only company to appear on every IA40 list. Featured today on the NYSE floor alongside fellow #IA40 honorees 📸

🆕 We've been busy shipping new features! Nick Karpov and Holly Smith are here to unpack what’s new and how it all fits together. In the latest Over Architected, they break down the newest Databricks updates, show how they work, and explore how they all fit together in one architecture. Covered in this episode: • Lakehouse//RT • Omnigent • Managed agent memory • Open Sharing • Managed disaster recovery • New Lakeflow connectors • LTAP • New model support Watch the full episode: https://lnkd.in/g3pX3PZy

Genie is changing how teams get answers from their data. In this DataFramed conversation, Databricks Co-Founder and Chief Architect Reynold Xin shares how he now uses Genie to investigate product metrics directly, and why that experience depends on the right data model, ontology, and governance underneath it. He also digs into Lakebase Postgres, LTAP, and how databases are evolving as agents become a primary persona. https://lnkd.in/g-nqzd7p

GLM 5.3 is available as a day 0 release on Databricks! GLM 5.3 is the smartest open-weight model available today and joins GLM 5.3 Flash and 30+ open-source and frontier models available natively on Databricks. We host GLM 5.3 on secure GPUs owned by Databricks so you can use the most intelligent models with your most secure data. With Unity Gateway, you can connect GLM 5.3 to your coding agents and get smart routing, centralized cost controls, and observability in a single pane of glass. https://lnkd.in/gyG8433E

Your data. Your context. Real outcomes. Attend #DataAIWorldTour 2026 to learn how organizations are grounding AI in enterprise data and business context. Hear the latest product announcements and customer insights, join technical breakouts, and explore practical ways to build and run AI apps, agents and analytics at scale. Find a city near you: https://lnkd.in/gxCdJEgu

Our partners are bringing a new wave of production-ready, industry-specific solutions to Lakebase. Built across financial services, manufacturing and energy, retail, healthcare, communications, and the public sector, these solutions pair industry expertise with a governed foundation for operational data, analytics, and AI. Use cases include: • Claims adjudication and fraud detection • Predictive maintenance and grid intelligence • Dynamic pricing and stockout detection • Prior authorization and behavioral health • Emergency response reporting Explore the partner solutions: https://lnkd.in/g3cCg5dh

Z.ai's latest open-weight model, GLM 5.3 Flash, is now available on Databricks! GLM 5.3 Flash joins 30+ open-source and frontier models on Databricks. On Databricks’ OfficeQA Pro v2 benchmark, it delivers 10% higher quality than GLM-5.2 at just one-tenth the cost, pushing out the quality-cost-pareto frontier. With new multimodal support, it can now read figures and verify web pages at coding. Run GLM 5.3 Flash where your data already lives — governed, secure, and ready for custom AI apps and agents. Control access, spend, and observability across all your AI with Unity Gateway. https://lnkd.in/gyG8433E

We asked a frontier agent to count the local maxima on a chart. It looked at the image, thought for 50 seconds, and got it wrong: 17 instead of 18. Then we gave Databricks Genie the same chart, but first extracted its values into structured JSON with ai_parse_document. Genie got the right answer. This is the core problem with agents and enterprise documents: the most important numbers often live inside charts, and a caption alone can't capture them. Agents can only search text, so they either retrieve the wrong page, or the right page without enough detail to answer. We tested a fix: parse charts into structured JSON at ingestion, index it with ai_prep_search, and let a lightweight 300M-parameter text embedding model do the retrieval. Across two chart-heavy benchmarks, this approach beat four multimodal embedding models up to 10x its size. Chart-JSON enrichment will power Genie One to answer questions over PDFs for databricks customers. It will also be available in ai_parse_document soon. Pair it with ai_prep_search to build agentic retrieval systems that actually get the numbers right. https://lnkd.in/gaT4byAS