Introducing the Nebius AI Builder Program: build AI systems you own on an open, independent stack

Today, we’re launching the Nebius AI Builder Program. It gives builders what they need to build and scale across the open AI ecosystem: credits across the stack, cookbooks, education and certifications, expert office hours, and a community of AI builders.

AI is no longer just a model you call. It’s a system you build. And the builders, not a few closed labs, will decide what it becomes. That system is a set of primitives working together: open models and fast inference at the foundation, with orchestration, retrieval, observability, evaluation, and post-training on top. Builders shouldn’t have to inherit all of those decisions from a handful of labs.

Much of the progress in AI is happening in the open now. AI is built by many teams, tested by the community, and improved by anyone with a better idea. That gives builders the choice to run the latest open models, pick the tools that fit, and change their mind later without rebuilding around one vendor.

What’s been missing is one place that puts it all within reach — the models, the tools, working code, and people who can help when you’re stuck. That’s why we launched the Nebius AI Builder Program

Open primitives: a stack you ownOpen primitives: a stack you own

Primitives are the building blocks that an AI system is made of: models, inference, orchestration, retrieval, evaluation, and coding agents. Builders can assemble these components into a stack they control, choosing open models and tools from across the ecosystem.

Say you’re building a support agent for a B2B product. You run Nemotron on Token Factory. It calls Tavily for live docs. When a ticket reports a bug, it’s automatically handed off to Devin or OpenHands to draft the fix. Then you, the human, close the loop. LangSmith evals show where the agent went wrong. Toloka puts human labels on those cases and Prime Intellect post-trains the model on them. With each cycle, the next version gets better at understanding your product and your customers.

Different tools work as one system, and you choose every one of them. There’s no lock-in. When a better open model ships, you point the agent at it, rerun your evals, and keep going. That’s what building on open primitives means in practice: the system gets better over time, and the decisions stay with you.

The AI Builder Program brings these primitives together in one place with credits and working code attached.

What you get with the Nebius AI Builder ProgramWhat you get with the Nebius AI Builder Program

The program is free. Here’s what members get:

• Credits across the stack. $400+ in credits and discounts from Nebius Token Factory, LangChain, Toloka, Tavily, and more.

• Cookbooks and working examples. Start from working code instead of a blank repo. Explore the cookbooks.

• Free courses. Nebius Academy and NVIDIA teach you how to build production AI agents with open models and developer tools, with more courses on the way for AI engineers and systems builders.

• Discounted certifications. Nebius certification at member pricing ($1).

• Office hours with expert engineers. Regular sessions to learn about new features, debug, tune performance, and get unblocked.

• A community that’s there when you start and as you grow. Join fellow builders on Discord and at live events around the world to share what you’re working on, get feedback, discover new tools and best practices, and connect directly with Nebius engineers. As your project matures, the community becomes a launchpad through hackathons, events, and opportunities to showcase what you’re building across Nebius and partner channels.

An independent ecosystem, assembled around your workloadAn independent ecosystem, assembled around your workload

Behind those benefits is an ecosystem, not a single vendor. At the heart of the stack are the models, compute, and inference: Nebius hosts the leading open models and provides the infrastructure to serve, train, and fine-tune them. Around that core, we’ve assembled the independent companies that complete the system.

The program launches with partners across every layer of the stack:

• Open models: NVIDIA Nemotron, Qwen, MiniMax
• Harness & coding agents: Cognition, OpenHands
• Agent frameworks: LangChain
• Agent search infra: Tavily
• Agent infrastructure: Composio
• Data, evaluation, and observability: Toloka, LangSmith
• Post-training and RL: Prime Intellect, Hugging Face OpenEnv

Come buildCome build

Ecosystems like this aren’t assembled from the top down. Builders assemble them, one system at a time, learning from each other as they go. That’s why the developer community is at the center of the AI Builder Program.

Bring what you’re building. Try the newest models the day they drop, start from working code, and compare notes with Nebius engineers, partner engineers, and fellow builders working on the same problems.

From your first agent to production inference, there’s always another turn of the loop. The stack keeps moving because builders keep building. Come be one of them.

Join the Nebius AI Builder Program