Databricks AI (formerly Mosaic AI)
Build, serve and govern AI agents next to your lakehouse data
Overview
Databricks AI, formerly marketed as Mosaic AI, is the AI layer of the Databricks platform: Agent Bricks, Model Serving and Foundation Model APIs, AI Search (formerly Vector Search), Unity Gateway governance, managed MLflow and agent evaluation. It grew out of Databricks' 2023 acquisition of MosaicML and is billed by consumption in Databricks Units.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Serverless-only, quota-limited workspace
- Build and evaluate AI agents
- No GPU serving endpoints or provisioned throughput
- One AI Search endpoint
- Full platform for business evaluation
- No credit card required
- Serve any model, feature or agent
- Autoscaling, billed on compute nodes used
- T4 at 10.48 DBUs/hour
- H100 at 100 DBUs/hour per GPU
- Committed-use discounts
- Custom requirements
What you can produce with Databricks AI (formerly Mosaic AI)
- AI agents built on enterprise data with Agent Bricks and the Agent Framework
- Model Serving endpoints for agents, generative and classical ML models
- Pay-per-token Foundation Model APIs for models such as GPT, Claude and Gemini
- AI Search vector indexes synced to Delta tables
- Unity Gateway controls on model, agent and MCP access, guardrails and spend
- Agent evaluation with AI judges and managed MLflow tracking
Databricks AI makes sense when your data already lives in Databricks: agents, model serving, AI Search and Unity Gateway governance all work against Unity Catalog without copying data to another platform. Agent Bricks and pay-per-token Foundation Model APIs cover most generative-AI needs.
Expect churn and metered costs. The Mosaic AI name is gone, fine-tuning has reached end of life, and two Agent Bricks products have closed to new customers within about a year. Serving starts at $0.07 per DBU, but GPU workloads add up quickly. Strong for existing Databricks teams; a heavy platform commitment for anyone else.
What we found
Databricks no longer markets anything as Mosaic AI. The name came from MosaicML, the generative-AI training company Databricks bought for about $1.3 billion in a deal completed in July 2023, and for two years it covered Databricks' model training, serving, vector search and agent tooling. Today the product page is simply Databricks AI, the documentation and pricing pages do not use the word Mosaic, and even the June 2024 press release that launched the Mosaic AI Agent Framework has been reworded to say Databricks instead.
Several parts were renamed along the way. Vector Search is now AI Search, the AI Gateway has become Unity Gateway, a control plane for which models, agents and MCP servers a team may use and how much it may spend, and Agent Bricks, launched in June 2025, is the headline for building agents on company data. Model Serving, Foundation Model APIs with pay-per-token access to models such as GPT, Claude and Gemini, managed MLflow and agent evaluation remain. Two things have gone: Foundation Model Fine-tuning has reached end of life, with Databricks pointing users to its serverless GPU AI Runtime instead, and the Agent Bricks Knowledge Assistant and Supervisor Agent are closed to new customers in favour of Genie Agents. Earlier versions of this page called fine-tuning the standout feature.
Pricing is consumption-based in Databricks Units. On the Premium tier in AWS US East, CPU and GPU model serving are both $0.07 per DBU, with GPU instances burning from about 10 DBUs an hour for a T4 to 800 for eight H100s. A two-week trial carries up to $400 of usage, and a Free Edition is available for learning, but it allows no commercial use, offers no GPU serving and lets Databricks train on what users put in.
The case for it is still data gravity: models, agents and governance sit next to data already in Unity Catalog. For teams outside Databricks, it is a platform decision, not an AI add-on.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Databricks AI (formerly Mosaic AI)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Databricks AI (formerly Mosaic AI) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Databricks AI (formerly Mosaic AI) directly →
Spotted something out of date? Suggest an update →
Alternatives to Databricks AI (formerly Mosaic AI)
More in Code & Dev Tools
