Over the past year, innovation in foundation models has advanced faster than anyone imagined. Today, the world’s top models come from a growing number of AI research labs, each excelling in different tasks, modalities, and price tiers. To balance quality, latency, and cost, enterprises are increasingly combining models to power their agents.

But that variety creates complexity: every model has its own API, authentication, and governance layer that teams must wire up and maintain.

As part of Week of Agents, Databricks is unifying the world’s leading foundation models and making them natively available inside the Databricks Data Intelligence Platform. Now, organizations can combine the strengths of OpenAI’s GPT-5, Google’s Gemini, Anthropic’s Claude, and Meta’s Llama with their own data to build agents that truly understand their business.

Unified access to the world’s best models

The world’s top foundation models, including OpenAI’s GPT-5, Google’s Gemini 2.5, and Anthropic’s Claude Sonnet 4.5, are now available natively inside Databricks. Through the Foundation Model APIs, they’re instantly accessible under the same authentication, security, and governance controls that already protect enterprise data.

“In aviation, speed and adaptability are everything, and the same goes for AI. With Databricks, Virgin Atlantic can securely run and govern models like OpenAI’s GPT-5, Claude Sonnet, Llama, and Gemma through a single platform. We can switch providers as needs change, manage costs through the AI Gateway, and still meet strict compliance requirements. That flexibility helps us make faster decisions, improve operations, and deliver better experiences for our customers and crew.”— Richard Masters, VP Data & AI, Virgin Atlantic

Each model family brings unique strengths, from reasoning power to speed to cost efficiency. On Databricks, customers can move fluidly along the curve of quality, latency, and cost, selecting the best model for every workload for both real-time and batch use cases.

With all these models in one place, developers can easily compare them side by side in the AI Playground, integrate them into applications through the REST API, or apply them at scale in SQL using ai_query, accelerating experimentation and bringing AI closer to production.

Govern every model with one control plane

As enterprises adopt multiple model families, platform teams face mounting complexity. Each provider introduces its own access rules, monitoring, and billing, creating governance sprawl and operational overhead.

Databricks AI Gateway removes that friction by serving as the control layer that makes multi-model AI enterprise-ready. Governance, observability, and usage tracking are unified across GPT-5, Gemini, Claude, and Llama, all within the Databricks Data Intelligence Platform. Platform teams gain a single view of usage, spend, and performance while Gateway enforces consistent permissions, guardrails, and rate limits, without the need for extra integrations.

Billing and capacity management are equally simple. With unified billing and shared reservations, enterprises can plan more predictably, manage costs effectively, and scale responsibly. Together, these capabilities help organizations stay at the frontier of AI quality while making governance sustainable at scale.

AI Gateway provides a unified control plane for all generative AI models across our enterprise, centralizing observability, cost governance, and capacity controls. We enforce governance requirements through input/output guardrails that prevent accidental exposure of PII to LLMs by masking sensitive data. Ultimately, AI Gateway is how we scale GenAI at The Standard, empowering confident innovation across the organization.Porter Orr, AVP Artificial Intelligence & Automation, The Standard

Connect models to your data in one platform

Now that leading models are available natively inside Databricks, the real advantage begins when those models meet your data. On Databricks, models and data live side by side, giving every model call the context it needs to deliver answers that truly reflect your business. This transforms generic models into intelligent agents that reason over real information rather than isolated prompts.

Agent Bricks makes this possible. It connects GPT-5, Claude Sonnet, Gemini 2.5 Pro, and other leading models to enterprise data in the Lakehouse, vectorized knowledge through Vector Search, and external systems via the Model Context Protocol (MCP). Building a declarative agent automates the complexity of selecting and optimizing the best combination of models, prompts, and data sources for each use case, and continuously improving performance through built-in observability and governance.

“Databricks gives us a secure, governed foundation to run multiple models, like GPT-5, Claude Sonnet, and Llama, and switch providers as our needs evolve, all while keeping costs in check. That flexibility lets us automate review moderation and improve content quality faster, so car shoppers get trusted insights sooner.” - Greg Rokita, Vice President of Technology, Edmunds

Start building

Access every leading model securely in seconds. **Open the Databricks **AI Playground and explore them side by side with your own prompts.

• Join us on 11/11 for “The Future of AI: Build Agents that Work” with Sam Altman and Ali Ghodsi.
• Read more about Week of Agents
• **For developers: **use theFoundation Model APIs to start building.
• **For platform teams: **set upAI Gateway for enterprise governance.