TensorZero Raises $7.3M Seed Round to Build an Open-Source Stack for Industrial-Grade LLM Applications

TensorZero, an open-source stack for industrial-grade LLM applications, led by Gabriel Bianconi, Viraj Mehta, and others, has raised $7.3 million in a seed round led by FirstMark, with participation from Bessemer, Bedrock, DRW, Coalition, and dozens of strategic angels.

Despite the hype, most companies building LLM applications still rely on a patchwork of early tools that don’t meet enterprise-grade cognitive or infrastructure demands. TensorZero solves this by offering a suite of production-ready components that work together seamlessly in a self-reinforcing loop—right out of the box.

TensorZero’s open-source stack unifies LLM gateway, observability, optimization, and evaluation into modular components that work together out of the box. It enables a principled feedback loop tied to KPIs, portable across model providers, and built for long-term advantage.

By automating workflows like variant suggestion, backtesting, and A/B testing, it lets engineers focus on high-level decisions while handling the low-level complexity behind the scenes.

TensorZero began in January 2024, with an early pilot in healthcare voice agents leading to its open-source launch in September. Since then, it has become the #1 trending GitHub repository globally, with contributions from developers worldwide. Today, it powers LLM products at leading startups and enterprises, including one of Europe’s largest banks.