Databricks Model Serving
A managed endpoint service for deploying and governing classical ML models, generative models and agents from one interface.
Find alternatives to Databricks Model Serving
What it does
- Model inference
- Model hosting
Features 6
- Simplified deploymentDeploy any model type, from pretrained open source to custom models, on CPUs and GPUs with automated container builds.
- Unified management for all modelsManage custom ML models, Databricks-hosted foundation models and externally hosted foundation models together.
- Batch inference with AI FunctionsServerless AI inference on large datasets across data types and models, integrated with SQL, Notebooks and Workflows.
- Built-in governance with Unity GatewayEnforces permissions, monitors model quality, sets rate limits and tracks lineage across hosted models.
- Fine-tuned and RAG model hostingHosts classical ML and generative AI models augmented with RAG or fine-tuned on enterprise data.
- Pay-per-token or provisioned throughputServes models as a low-latency serverless API with CPU and GPU support that scales from zero.
How it compares
-
Hugging Face Hub
Databricks overlaps with Hugging Face, where enterprises want integrated data-to-model pipelines
Research report · 19 Sep 2026
Sources
- Pricing
- Description
- Constraints
-
Derived from the security facts on this page, from the same source.
- Security
-
Platform-wide: databricks.com/trust (+/compliance, /privacy, /security-features) and docs.databricks.com security overview. SOC 2 Type II report stated; ISO certifications mentioned without numbers so not recorded; DPA listed under privacy resources. FedRAMP High only for a separate GovCloud offering, not recorded. Training/ZDR/residency not stated on these pages.
- Sold within
-
Its own per-plan rate card (CPU and GPU Serving $0.070 / DBU, GPU DBU/hour table) billed on a platform SKU: "Usage of these services will be appear on your bill under the Serverless Real-time Inference SKU (or under the Model Serving SKU on SAP Databricks)." [sic] The page also says "Pay as you go with a 14-day free trial or contact us for committed-use discounts or custom requirements."
- Interfaces
-
Official documentation · 12 Oct 2026
Page: 'accessible in a unified user interface and API'.
- Intended for
-
"Databricks Model Serving provides enterprises with a robust solution for deploying classical ML models, Generative AI models, and AI agents"
Also from Databricks
-
Agent Bricks Agent platform
A control plane for building, evaluating, governing and monitoring AI agents across proprietary and open-source models, with native MCP support.
-
Databricks AI Search Data service
A managed hybrid semantic, keyword and vector search service (formerly Mosaic AI Vector Search) with automatic data sync and Unity Catalog governance.
-
Databricks Model Training Developer tool
A managed service for fine-tuning open-source LLMs or training custom models on enterprise data using dedicated GPU infrastructure.
-
Databricks AI Platform
Databricks' tooling for building, fine-tuning, serving and evaluating models and agents on lakehouse data.
-
Databricks Data + AI Platform Platform
Lakehouse platform for storing, governing, querying and sharing enterprise data.
-
AI/BI Application
AI-native business-intelligence product comprising AI/BI Dashboards (AI-assisted dashboard/visualization creation) and Genie Spaces (conversational natural-language exploration of data), built into the Databricks Data + AI Platform with Unity Catalog governance and no per-seat licensing.
-
Genie One Application
AI coworker (the evolution of the earlier Databricks Assistant/Genie) that lets business users ask questions, take action and drive outcomes over enterprise data via natural language, integrating with Slack, Teams, Jira, Google Drive and Salesforce, with mobile apps.
-
Unity Gateway Infrastructure service
Enterprise AI gateway providing centralized cost tracking/budgets, model access (Claude, GPT, Gemini, Grok and others), security/governance (access policies, PII/PHI filtering, audit trails), smart routing and observability across an organization's AI systems.
-
Document Intelligence Data service
Document Intelligence uses AI functions to parse, extract, and classify enterprise documents.
-
Databricks Feature Store Data service
Databricks Feature Store provides a governed layer for building, sharing, and serving machine learning features.
-
Genie Agents AI agent
Genie Agents answer natural-language data questions with SQL queries, result tables, and visualizations.
-
AI/BI Dashboards Application
AI/BI Dashboards provides AI-assisted authoring, data visualization and forecasting.
-
AI Runtime Infrastructure service
AI Runtime provides serverless GPU compute for training, fine-tuning and batch inference of custom models.
-
Genie Code AI agent
Autonomous AI agent for data teams inside the Databricks workspace: plans and runs multistep data engineering, data science, machine-learning and dashboard work in notebooks, the SQL editor and pipelines, grounded in Unity Catalog metadata; also engineers features and trains and evaluates models.
-
Databricks CustomerLake Application
Agentic customer data platform embedded in Databricks: builds Customer 360 profiles with agentic identity resolution, syncs audiences to martech and adtech tools, and runs marketing agents that plan and personalise campaigns; also answers natural-language questions about customer data.
-
Lakewatch Platform
Agentic SIEM built on the Databricks security lakehouse: ingests and normalises security telemetry to OCSF, manages detections as code, and runs agents that author rules, triage alerts and hunt threats.
-
Databricks AutoML Developer tool
Automated machine learning that trains and evaluates classification, regression and forecasting models on a dataset, and writes an editable notebook with the training code for every trial.
-
Managed MLflow Developer tool
Managed MLflow on Databricks: experiment tracking, tracing, evaluation and model management for classical machine learning, deep learning, generative AI applications and agents, governed through Unity Catalog.
Open Databricks in the directory
Something wrong here? Send a correction — quote this product id: databricks-model-serving.