Building a compliance audit agent using Nebius Agents Blueprint
Using Nebius Agents Blueprint, we built a regulatory compliance audit agent and ran the same audit task through four configurations — from a $470 prototype to a production system at 72% lower cost with 1.00 recall. This post covers what changed at each step, what each configuration revealed about the one before it, and the production failures that only became visible once the operational architecture existed to find them.
Introducing the Nebius Agents Blueprint: open architecture for production-ready AI agents
Today we’re introducing the Nebius Agents Blueprint, an open reference architecture for building, operating, and continuously improving AI agents in production. This post covers the six-component composable stack — inference, orchestration, retrieval, grounding, observability, and simulation — and the case study behind it: a compliance audit agent that achieved 72% lower cost and 20% higher precision over a GPT-based prototype by improving the system, not the model.
NVIDIA retail AI blueprints, now running on Nebius
Nebius has collaborated with NVIDIA to bring two retail AI blueprints to production on Nebius infrastructure: the NVIDIA Agentic Commerce Blueprint and the NVIDIA Retail Catalog Enrichment Blueprint for automated product content. Both are open-source reference architectures built on NVIDIA NIMs that retailers and developers can customize and deploy today with a 1-click deployment on Nebius AI Cloud.
Building transaction foundation models on Nebius AI Cloud
Today, we’re exploring how transaction foundation models move from developer examples to production systems. Using NVIDIA’s TFM blueprint and Revolut’s PRAGMA model, we show how Nebius AI Cloud supports the full lifecycle — from GPU-accelerated data preparation and multi-node training to managed inference on Token Factory.
Run physical AI workflows, not glue code
The Nebius Physical AI Workbench turns NVIDIA Cosmos 3, NVIDIA Isaac Sim, NVIDIA Isaac GR00T and other Physical AI tools into composable building blocks that agents can wire together. We are building it in the open, and it is available now on GitHub.
Nebius and Tavily: Bringing agentic search into the production AI stack
Nebius has acquired Tavily to bring agentic search into the Nebius AI Cloud platform. In this post, we explain how Tavily adds real-time web access to the Token Factory stack, helping developers build production AI agents that can reason, verify information and act on live context instead of relying only on static model knowledge.
Data Lab: Your best dataset is already in your logs
Today, we’re launching Data Lab in Nebius Token Factory. It is a new workspace for turning production logs and existing datasets into reusable training data for post-training workflows. Data Lab helps teams explore inference logs, curate datasets and move directly into model iteration without rebuilding pipelines or copying production data across environments.
The AI cloud will be won at the software layer
Yesterday we announced that Clarifai, a core AI engineering and research team, is joining Nebius. This follows the recently-announced agreement to acquire Eigen AI. Both are deliberate moves on the same bet: that the AI infrastructure opportunity will be decided at the software layer.
Game on! UFB and Nebius power the future of robotics
We are thrilled to announce that Nebius is the official technology partner for Season 1 of Ultimate Fighting Bots (UFB)’s Humanoid Sports League — the world’s first multi-format competition platform for physical AI. The two companies will pioneer the use robots by coupling UFB’s growing platform for humanoids with Nebius’s cloud solution for physical AI.
Nebius and LangChain partner to power production-grade AI agents on open models
Nebius and LangChain have partnered to integrate Nebius Token Factory with LangChain’s Deep Agents. The integration, combined with LangChain’s existing Tavily integration, gives teams building on LangChain a direct path to run agent workloads on production-grade AI infrastructure with open-source models, dedicated endpoints, real-time search and full control over cost and data.
[Security Advisory] CVE-2026-43284, CVE-2026-43500: “DirtyFrag” Linux kernel local privilege escalation — mitigation required
DirtyFrag (CVE-2026-43284 and CVE-2026-43500) is a pair of Linux kernel vulnerabilities that can allow local privilege escalation on affected systems. In this advisory, we explain which Nebius services may be impacted, what mitigations are currently recommended and how to secure compute instances and Managed Kubernetes environments.
[Security Advisory] CVE-2026-31431: Copy-fail vulnerability requires immediate mitigation on Nebius instances
A critical Linux vulnerability known as Copy-Fail (CVE-2026-31431) affects systems running the algif_aead kernel module and may impact Nebius compute instances, Kubernetes clusters and containerized workloads. In this post, we explain which services are affected, what mitigations have already been applied by Nebius and what actions customers should take to secure existing environments.
