Labelbox has raised more than $100 million in Series E financing to expand its platform for reinforcement learning, AI model evaluation and enterprise agent deployment. The round, comprising equity and a credit facility, was led by Infinitum Partners, with participation from Wipro Ventures, Andreessen Horowitz, B Capital, Tau Ventures, Snowpoint Ventures and other investors. The company is investing in two products: Horizon, which provides training and evaluation infrastructure for frontier AI models, and Recursion Managed Agents, designed to run AI agents inside enterprise applications. The financing comes as AI developers look beyond model training toward systems that can execute complex tasks and improve through feedback.
Labelbox targets the next phase of AI development
AI companies are increasingly focused on what happens after a model has been trained. Beyond generating text or answering questions, advanced systems need to handle multi-step tasks, use tools, respond to feedback and produce reliable results in real-world settings.
Labelbox is positioning its platform around that transition, combining infrastructure for reinforcement learning with managed AI agents for business workflows. The company said it serves more than 400 customers, including over 90% of leading U.S. AI labs. It also reported reaching profitability in 2026 while growing more than fivefold year over year. These are company-reported figures; the announcement did not disclose revenue, the definition of profitability or the precise period used to calculate growth.
The Series E financing exceeds $100 million and includes both equity and a credit facility, meaning the headline amount should not be interpreted as entirely new equity capital. Infinitum Partners led the round, joined by Wipro Ventures, Andreessen Horowitz (a16z), B Capital, Tau Ventures, Snowpoint Ventures and other investors.
Labelbox said the funding will support research and engineering for its two core products and expand the teams responsible for bringing them to customers. The company did not disclose a valuation or provide a breakdown of the financing between equity and credit.
Horizon brings reinforcement learning infrastructure to AI labs
Labelbox’s Horizon product targets frontier AI developers that need environments in which models can practice complex tasks, receive feedback and be evaluated.
Reinforcement learning trains a system using signals about the quality of its actions or outcomes. In modern AI development, it can complement other training techniques by helping models improve their performance on tasks that require multiple steps, tool use or adherence to defined objectives.
Horizon provides simulated environments, expert feedback and evaluation infrastructure for tasks including coding and research. Rather than relying solely on static training examples, these environments allow models to attempt work, receive performance signals and repeat the process.
Such infrastructure can be important when developers want to assess whether a model can complete a task, not simply produce a plausible answer. Evaluation environments can also help teams compare model behavior across different conditions and identify failure modes before deploying systems more broadly.
Labelbox said it works with leading AI labs and hyperscalers on post-training programs. It did not identify those customers in the announcement or disclose comparative performance results showing how Horizon changes model capabilities.
The wider opportunity is tied to the complexity of advanced AI development. As models are expected to operate across software tools, coding environments and research workflows, developers need repeatable ways to test and refine performance. However, the usefulness of reinforcement learning depends on the quality of the environments, the feedback signals and the evaluations used to measure progress.
Recursion Managed Agents targets enterprise execution
The company’s second product, Recursion Managed Agents, addresses a different stage of the AI lifecycle: putting agents to work inside business applications.
The platform is designed to deploy groups of agents that can start tasks in response to schedules, alerts, support tickets and other triggers. Agents can coordinate across tools, execute workflows and have their outputs assessed against predefined standards.
Labelbox says the agents can use information from previous runs to inform their memory and improve subsequent work. That approach aims to make agents more useful over repeated tasks rather than treating every execution as an isolated interaction.
The company said it already uses more than 50 agents internally across engineering, research, product, sales, marketing and finance. It opened Recursion Managed Agents to enterprise customers earlier in October 2026, according to its announcement.
Internal use provides an initial example of how Labelbox applies its own technology, but it does not independently establish how the product performs across customer environments. Enterprises will still need to evaluate reliability, access permissions, security, auditability and the ability to intervene when agents make mistakes.
Integrating agents into existing applications may reduce the need to replace established software systems. Yet deployment at scale also raises questions about who owns agent actions, how sensitive data is handled and whether automated workflows can be monitored and reversed when necessary.
One platform for model improvement and deployment
Labelbox’s strategy connects model development with operational use. Horizon supports training and evaluating AI systems, while Recursion Managed Agents focuses on executing business tasks and using feedback from prior runs.
The two products serve different customer needs, but they reflect a shared industry goal: making AI systems more capable of completing work with less step-by-step human direction. A platform spanning both stages could appeal to organizations seeking continuity between model improvement and enterprise deployment.
CEO and co-founder Manu Sharma said the funding would enable the company to advance sandbox infrastructure, agent systems and reinforcement learning post-training. The announcement did not specify how much capital will go to each product or provide a timetable for major technical releases.
Competition in this market includes a broad mix of model developers, AI infrastructure providers, evaluation platforms and enterprise automation vendors. Labelbox’s differentiation will depend on how effectively its training environments, evaluation systems and managed agents address customer needs, rather than simply offering multiple capabilities under one umbrella.
The financing gives the company additional resources to pursue that strategy. Its next challenge will be demonstrating that systems trained or evaluated in controlled environments can perform reliably in production, and that agents can deliver consistent results across the varied tools and processes used by enterprises.
Market Landscape
AI infrastructure is expanding beyond model training into post-training, reinforcement learning, evaluation and production agent management. Developers need environments that can test complex behavior, while enterprises need systems that execute workflows with appropriate oversight. Labelbox is targeting both sides of this market through Horizon and Recursion Managed Agents. Its growth prospects will depend on customer adoption, measurable model improvements, reliable agent performance and its ability to differentiate itself in an increasingly crowded AI infrastructure market.
Top Insights
- More than $100 million raised: Labelbox’s Series E includes equity and a credit facility, with Infinitum Partners leading the round and several existing and new investors participating.
- Two complementary products: Horizon focuses on reinforcement learning and evaluation for AI labs, while Recursion Managed Agents targets enterprise workflow execution.
- Reported growth and profitability: Labelbox says it became profitable in 2026 while growing more than fivefold year over year, without disclosing detailed financial figures.
- Enterprise agent deployment: The company says it uses more than 50 agents internally and recently opened its managed-agent product to enterprise customers.
- Execution remains the test: Commercial success will depend on measurable model improvements, dependable agent workflows, security controls and sustained customer adoption.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
