Scale AI

Today, we’re announcing Scale has closed $1B of financing at a $13.8B valuation, led by existing investor Accel. For 8 years, Scale has been the leading AI data foundry helping fuel the most exciting advancements in AI, including autonomous vehicles, defense applications, and generative AI. With today’s funding, we’re moving into the next phase of our journey: accelerating the abundance of frontier data to pave the road to Artificial General Intelligence (AGI). “Our vision is one of data abundance, where we have the means of production to continue scaling frontier LLMs many more orders of magnitude. We should not be data-constrained in getting to GPT-10.” - Alexandr Wang, CEO and founder of Scale AI. This new funding also enables Scale to build upon our prior model evaluation work with enterprise customers, the U.S. Department of Defense, and collaboration with the White House to deepen our capabilities and offerings for both public and private evaluations. There’s a lot left to do. If this challenge excites you, join us: https://scale.com/careers Read the full announcement: https://lnkd.in/gVBhaPZ5

Last week we hosted the first Scale St. Louis Summit at our AI Center in downtown STL! We brought together leaders from industry, higher education institutions, and the US government to discuss St. Louis’s potential to become a defense tech hub and the national security contributions made by the city. Thank you to keynote speaker, Congressman Wesley Bell, Senator Eric Schmitt, and panelists Kristin Sobelik, Maggie Kost, and Andy Dearing. Learn more about the St. Louis AI Center here: https://scale.com/stl

Qwen3 is live on the SEAL leaderboards and notably stands out among other open source models. View the leaderboards: scale.com/leaderboards

Today, our CEO Alexandr Wang joined the Center for Strategic and International Studies (CSIS) Wadhwani Center for AI and Advanced Technologies for a conversation on the future of AI leadership and national security. Moderated by Gregory C. Allen, Director of the Wadhwani AI Center, Alex shared Scale’s perspective on AI’s critical role in strengthening U.S. national security, advancing safe and responsible AI development, and our work with partners like the Department of Defense and the AI Safety Institute to help set global standards. You can watch the full event here: https://lnkd.in/eemWz_Ck

Your enterprise agents likely aren’t failing because the model’s wrong—they fail because the system around it wasn’t built for the real world. But these considerations aren’t insurmountable. In the latest episode of Human in the Loop, our Head of Enterprise ML Sam Denton, Head of Product for Scale GenAI Platform Clemens Viernickel, and Head of Engineering, Enterprise AI Felix Su dive into the core technical challenges of making agentic systems actually work in enterprise environments—across real data, real constraints, and real complexity. They discuss: 👉 Why expert feedback is essential for effective agents 👉 How to scaffold workflows to minimize compounding error rates 👉 Why privacy, compliance, and access controls must be built in, not bolted on 👉 How to architect systems that adapt to constantly evolving data sources Whether you’re already building enterprise-grade agents or looking to bridge the gap between agents you build on your local machine and enterprise-grade systems, this is the episode for you. Watch the full episode here: https://lnkd.in/gqMJ2iaz

Building agents is as much a product problem as an AI problem. Join us for a webinar walking through the detailed blueprint you need to follow to set up successful agents in your workplace. General Manager and Director, International Public Sector, at Scale, Sahil Bhaiwala walks through the framework and tactical instructions to build professional-grade agents, plus how you can evaluate them. When: April 30, 11:00 AM ET Register: https://lnkd.in/g9WMEYc3

The agent landscape is getting increasingly complex. What should enterprises pay attention to as they adopt agents in production? In the latest episode of Human in the Loop, our new video podcast series, our Head of Product, Enterprise Solutions, Ben Scharfstein, and Head of Engineering, Enterprise AI, Felix Su, dive into the current AI agent landscape and cover what’s important for enterprises to move beyond demos to real, reliable agentic systems. They dig into: 👉 What “agents” really are—and where most demos fall short 👉 The tools and frameworks gaining traction (and which ones actually matter) 👉 Why precision, observability, and tight product integration are non-negotiable for enterprise use 👉 How to avoid compounding errors and build systems that stay aligned over time If you’re building agents in the enterprise–or thinking about it–this is the conversation to start with. Listen here: https://lnkd.in/dvxm4pBs

Pushing the AI frontier means testing skill integration: how capable are agents at complex, multi-step workflows like those needed for research? OpenAI's PaperBench, released earlier this month, evaluates replicating papers without code, testing skill integration. Results show agents have non-trivial capabilities, but lag significantly behind human experts, primarily due to challenges in task execution. On the Scale blog, we explore how this vital benchmark pinpoints R&D needs for robust agentic AI and informs safety efforts. Read the full post for our take on PaperBench here: https://lnkd.in/g4nmmtX4

o3 and o4-mini have the lowest calibration error to date on Humanity’s Last Exam. We had to ask: were they actually better-calibrated? Our research team at Scale ran a deeper analysis and found the low calibration error scores are real, but not for the reasons you might think. Dive into the analysis: https://lnkd.in/gAgeq3ev

o3 and o4-mini are out today and already on the SEAL leaderboards! Spoiler alert: o3 dominates 🔥 o3: 1st (tie) - Humanity’s Last Exam 1st (tie) - MASK 1st - MultiChallenge 1st - EnigmaEval View the leaderboards: Humanity’s Last Exam: https://lnkd.in/gKSZptdg  MultiChallenge: https://lnkd.in/gZVmy6SW  VISTA: https://lnkd.in/g43qgC_R  MASK: https://lnkd.in/gZTUTvGe  EnigmaEval: https://lnkd.in/g6qEfqCS