Founders Fund, Pantera, and Franklin Templeton join Sentient's "Arena" to conduct stress tests on enterprise-level AI agents

Over the past two years, companies have been accelerating the integration of AI agents into real workflows: from customer service and backend operations to finance and compliance processes that require high-stakes decision-making. As these systems become more embedded in actual business operations, a new challenge is emerging: while agents can retrieve information, they often struggle to provide stable, explainable, and reproducible reasoning processes when tasks become “dirty,” multi-step, or high-risk.

Today, open-source AI lab Sentient officially launched Arena—a real-time, production-grade environment designed for thousands of AI developers worldwide to stress-test and iteratively improve various enterprise reasoning challenges. The initial participants include Founders Fund, Pantera, and Franklin Templeton, which manages over $1.5 trillion in assets—sending a clear signal: institutions are showing early and explicit interest in “structurally evaluating AI agents before deployment.”

“When companies apply AI agents to research, operations, and customer-facing workflows, the question is no longer whether these systems are powerful enough… but whether they are reliable in real-world scenarios,” said Julian Love, Partner at Franklin Templeton Digital Assets. Love added that structured environments like Arena will help the industry distinguish between “promising ideas” and “capable systems ready for production.”

Himanshu Tyagi, co-founder of Sentient, stated, “AI agents are no longer just experiments within enterprises; they are entering critical processes that impact customers, capital, and operational outcomes. This shift changes the evaluation criteria. A system that looks impressive in a demo isn’t enough. Companies need to know: in a production environment where failure is costly and trust is fragile, can the agent still reason reliably? They need comparability, reproducibility, and a way to track long-term improvements without relying on underlying models or tool stacks.”

Arena simulates the chaotic reality of enterprise workflows: incomplete information, long contexts, vague instructions, conflicting sources. It doesn’t just assess whether agents produce the “correct answer,” but records complete reasoning traces to help engineering teams identify failure points and verify whether improvements are effective over time.

This provides a neutral, vendor-agnostic benchmark for reasoning evaluation across models and tech stacks. Arena emphasizes production-level performance rather than demo results, enabling the development of verifiable, high-stakes capable agents that can be migrated to private data and internal tools.

In the first challenge, developers participating in Arena will focus on a core enterprise problem: document reasoning. AI agents must reason and compute over complex, unstructured data—an essential underpinning for financial analysis, root cause investigation, investment memos, customer service, and more.

Other initial participants include alphaXiv, Fireworks, OpenHands, and OpenRouter. As Arena expands to incorporate more tasks, industries, and model integrations, more participants are expected to join.

Recent surveys highlight the gap Arena aims to address: 85% of companies want to become “agentic enterprises,” nearly three-quarters plan to deploy autonomous agents, but fewer than a quarter have mature governance systems in place; many struggle to scale pilots into large-scale deployments. On average, companies are running about a dozen agents, often in isolated scenarios; many believe that without better orchestration and coordination, adding more agents will only increase complexity and reduce value.

“At OpenHands, we’ve always been eager to support developers using agents to solve real, practical problems,” said Graham Neubig, Chief Scientist and co-founder of OpenHands. “We’re also excited to support participants using the OpenHands Software Agent SDK to tackle these complex challenges.”

Alex Atallah, co-founder and CEO of OpenRouter, added, “Arena is exactly the kind of initiative that can push open-source AI forward—it allows researchers to compete, iterate, and innovate in an open environment. We look forward to deepening our collaboration with Sentient and providing infrastructure to make experiments faster, easier, and more scalable.”

Arena will launch globally, inviting thousands of AI developers to apply for the first limited cohort, with in-person events starting in San Francisco from March 2026.

About Sentient Labs

Sentient Labs is a leading research and product organization dedicated to advancing open-source AI. As the innovation engine under the Sentient Foundation, Sentient Labs conducts cutting-edge research in AI reasoning, alignment, and agent collaboration. It is the core developer of high-performance frameworks like ROMA and open-source models such as Dobby. Sentient’s mission is to move open-source AI from “experimental” to “essential.” By providing infrastructure to build powerful, composable agent systems, Sentient enables developers to commercialize open-source tools and achieve enterprise-level usability. Sentient is committed to making open-source the default standard for mission-critical AI operations worldwide.