Releasing Hopsworks 5.0 - Introducing the Coding Data and...

The team at Hopsworks is excited to announce Hopsworks 5.0, the unified and Sovereign Data and AI platform built around the Coding AI and Data Stack; a new paradigm where coding agents replace fragmented tooling to accelerate the development and maintenance of ML pipelines, data workflows, and AI applications and much more.

It marks a new chapter in how AI systems get built and maintained. For the first time, teams have a full coding agent and terminal embedded inside the platform itself, pre-loaded with Claude Code and Codex. There is no context switching as developers describe what they want to build, and the platform handles the rest from ingesting raw data to deploying production-ready dashboards and inference pipelines.

The release, generally available, is designed to help small and large teams alike build, iterate, and operate AI pipelines faster than ever before from batch and real-time systems to LLM-powered applications.

Key capabilities of Hopsworks 5.0:

Development Containers, Coding Agent and Terminal A terminal and development containers are built directly inside Hopsworks to allow agent to instruments the platform, giving developers a unified environment to build and operate the full ML stack; feature pipelines, training pipelines, inference pipelines, Streamlit applications, and analytics dashboards without leaving the platform. Claude Code and Codex come pre-installed, and multiple agent sessions can run in parallel, so developers can work across different parts of a project simultaneously. Developers describe what they want to build in plain language, and the agent writes, runs, and iterates on the code. Sessions are fully persistent across restarts, so work is never lost between coding sessions.

New User Interface A completely redesigned user interface that delivers 50% lower latency across the most common platform tasks and meant to simplify user experience across the whole platform, with the help of the Wizard interface.

The Wizard allows users to select any case, such as time series forecasting, auto research, and others , automatically driving through steps and a guided setup that delivers a production-ready pipeline with a coding agent.
Platform Intelligence An intelligent agent is embedded across key platform workflows. When connecting external data sources from Databricks, Snowflake, BigQuery, MongoDB, SAP HANA, and others, Platform Intelligence uses an LLM to generate human-readable column names and descriptions, infer primary keys and event time columns, and configure ingestion schedules automatically.

Analytics with Trino and Apache Superset Hopsworks now supports native SQL query capabilities powered by Trino and a full dashboarding layer via Apache Superset. Teams can run exploratory SQL queries, build and share dashboards, and export results as PDFs or PNGs without leaving Hopsworks. Batch inference pipelines can write predictions directly to feature groups, making results immediately queryable and visualisable.

Expanded Data Integration Hopsworks 5.0 introduces a significantly expanded set of data sources alongside two new ways to work with external data: mounting external tables without copying data, and ingesting data to Hopsworks using DLTHub on a configurable schedule.

Column-Level Access Control Column-level access control enables the sharing of selected columns from tables (feature groups) to teams in different projects. In ML, it is common for data owners to manage wide tables with some columns (features) used in many models, but other columns containing sensitive information. Column-level access control gives data owners, precise control over what columns from the (wide) tables are shared, with whom.

Model Registry Enhancements Models can now be imported directly from Hugging Face, including the model card documentation, making it straightforward to bring models into on-premise or air-gapped deployments. Deployed models benefit from Platform Intelligence for configuration, and full provenance linking training data, feature views, and model versions is tracked automatically.

Read more about out the Hopsworks 5.0 release notes here

"Hopsworks 5.0 is our most ambitious release yet. It is a big step in moving Hopsworks from a MLOps platform to an AI Lakehouse, with support now for dashboards at scale, a new query engine, Trino, and native Apache Iceberg support. But, Hopsworks 5.0 is not a me-too Lakehouse. It is a new class of AI Lakehouse powered by coding agents. The low-code and no-code tooling that dominates other Lakehouses is obsolete. Instead, we have added first class support for building and operating pipelines, agents, apps, and dashboards from a single Terminal UI (a dev container). All this as we continue to expand our lead in real-time AI, with more support for real-time transformations in RonDB." — Jim Dowling, CEO, Hopsworks

Availability

Hopsworks 5.0 is generally available. Hopsworks SaaS, which includes the full Coding Data Stack experience and the Wizard, is open for registration. Both on-premise and cloud deployments are supported for enterprise customers, with GPU and distributed training via Ray and Spark for large-scale workloads.

About Hopsworks

Hopsworks is the leading unified platform for building batch, real-time, and LLM-powered AI systems, enabling organisations to go from raw data to production-ready AI and Data products at scale. As a European-founded, sovereign AI platform, Hopsworks is purpose-built for organisations that require data sovereignty and regulatory compliance. At its core is a powerful AI Lakehouse, combining a best-in-class feature store, model registry, and vector database covering the full pipeline lifecycle from data ingestion to batch, real-time, and agentic inference.