Key facts

Founded
2018
HQ
San Francisco, United States
Employees
—
Funding stage
Series E · US$100m+
Further facts
  • Funding: Labelbox raised more than US$100m in Series E equity and credit on 14 October 2026, led by Infinitum Partners, at a US$1.25bn pre-money valuation.
  • Capital: Labelbox has raised US$189m in total and is profitable.
  • Reach: More than 400 customers build with Labelbox, including 90% of leading US AI labs.

Focus areas

Reinforcement learning Data labelling LLM evaluation AI agents Model training

About

Labelbox is a San Francisco AI company founded in 2018 that started with software for building machine-learning training data. Manu Sharma is co-founder and chief executive. Its Horizon business builds the reinforcement-learning environments and evaluations that frontier labs post-train models in, and its Recursion platform runs fleets of agents for enterprises. On 14 October 2026 it raised more than US$100m in Series E equity and credit led by Infinitum Partners, with Wipro Ventures, a16z, B Capital, Tau Ventures and Snowpoint Ventures, at a US$1.25bn pre-money valuation; it has raised US$189m in total and is profitable.

How they use AI

Horizon builds simulated work environments, expert-written graders and evaluations so frontier models learn by practising real tasks rather than copying answers. Recursion Managed Agents start on their own from schedules, alerts, commits or API calls, work across tools such as GitHub, Jira and Slack, and a grader agent scores every run. The best graded runs train specialist models, and Labelbox runs more than 50 of its own internal agents on Recursion.

Products and services

  • Horizon — RL environments and evaluations for frontier AI labs
  • Recursion — Managed agents that run graded enterprise work
  • Labelbox platform — Software for building expert training data

Key people

  • Manu Sharma — Co-founder and CEO

Investors