Labelbox has raised more than $100 million in Series E funding to grow its reinforcement learning tools and enterprise AI agents. The San Francisco company, best known for helping AI labs prepare training data, announced the round on October 14, 2026. The money is going toward two products that help AI models and agents improve by practicing real tasks.
Inside the Labelbox Series E Funding
The round mixes equity with a credit facility. Infinitum Partners led it, with Wipro Ventures, Andreessen Horowitz, B Capital, Tau Ventures, and Snowpoint Ventures also taking part.
According to coverage of a founder’s note from CEO Manu Sharma, the deal values Labelbox at $1.25 billion before the new money. Sharma also said access to computing power will be central to the company’s next chapter.
| Detail | What Labelbox reported |
|---|---|
| Amount raised | More than $100 million (equity plus credit) |
| Lead investor | Infinitum Partners |
| Pre-money valuation | $1.25 billion |
| Customers | More than 400 |
| Growth | More than 5x year over year, profitable in 2026 |
| Founded | 2018 |
Labelbox says it works with over 90% of leading US AI labs. Its customer list also includes cloud giants, startups, and Fortune 500 companies.
How Labelbox Trains AI Through Practice
The first product, Horizon, is built for labs that train frontier models. It creates simulated work environments where a model tries hard, multi-step tasks and gets scored on how it did.
The company says these simulated companies come filled with code, dashboards, spreadsheets, and documents. Each attempt is recorded and graded, and expert reviewers help teach the model which correct answers are actually better.
Why Reinforcement Learning Matters Now
Reinforcement learning lets a model improve by trying, getting feedback, and trying again. Labelbox argues this is how the most capable models now get better, and it is the main reason demand for its services has jumped.
Horizon covers areas like coding, computer use, research, voice, and cybersecurity. Labelbox also says it runs an expert network of more than 4 million people to supply human feedback.
Labelbox Brings Self-Improving AI Agents to Businesses
The second product, Recursion Managed Agents, opened to enterprise customers earlier in October. It lets companies run fleets of AI agents inside the apps their teams already use.
Agents can start work from a schedule, an alert, a ticket, or a chat message. A separate grader agent checks each run against a set of rules, and lessons from past runs are saved to the agent’s memory.
- Agents work across more than 100 business apps and any MCP server.
- Each tool is labeled as read-only, data-changing, or destructive.
- Login credentials stay in a vault, so the model never sees them.
- Pricing is set per agent session hour, with compute and model costs combined.
Labelbox says it uses more than 50 agents internally, across teams like engineering, sales, and finance. In one example, the company reported that the cost of its code review agent fell about 32% per review as the agent learned.
The pitch here is simple: an agent that remembers what went wrong last time should cost less and do better the next time.
What This Means for You
Most people will never use Labelbox directly, but its work shapes the chatbots and assistants you use every day. Here is why this round matters.
- Better-trained models: more practice environments for AI labs can mean fewer mistakes on complex tasks.
- Smarter workplace tools: agents that learn from each run could make business automation more reliable over time.
- Safety checks matter: features like tool permissions and credential vaults are worth looking for in any agent product.
- A crowded market: training data and agent tools are now one of the hottest areas for AI investors.
This round lands in a busy week for AI funding. Recent deals include Modal’s $750M raise for AI cloud sandboxes and Harmonic Security’s funding for safer workplace AI.
The value of training data has also driven odd deals, like AI labs that buy startup data. You can follow more stories on our AI news page.
Sources
This article is based on Labelbox’s announcement via PR Newswire and coverage from Unite.AI.
Conclusion
Labelbox has raised more than $100 million in a Series E led by Infinitum Partners, at a reported $1.25 billion pre-money valuation. The company says it is profitable, growing more than fivefold a year, and working with most leading US AI labs. The money will go to Horizon, its training environments for frontier models, and Recursion Managed Agents, its self-improving business agents. It is another sign that teaching AI to learn from real work has become one of the biggest bets in the industry.
