Labelbox
Labelbox builds RL environments and expert human data for frontier AI labs, plus a labeling platform with a free tier and Alignerr expert network
Overview
Labelbox started as a data labeling platform and now pitches itself as the place "where AI learns to do real work": it "builds the environments frontier labs train AI on, and the agents engineering and research teams hand their backlog to." The homepage leads with newer products (Recursion managed agents, Horizon with WorldSim, Terra) and the Alignerr expert network, with separate tracks for enterprises and frontier AI labs. The self-serve platform is still documented: Annotate (10+ built-in editors), Catalog (curation and search across 25+ cloud sources), Model (predictions and error analysis), Foundry (foundation models for RLHF preference data), team management, and an MCP-based connection for AI assistants. Tasks cover multimodal chat, LLM evaluation, computer vision, NLP, video, audio, coding and agent work, with expert labeling in 30+ languages. In October 2026 Labelbox raised more than $100M in a Series E at a $1.25B pre-money valuation, led by Infinitum Partners; the company says it is profitable and grew revenue more than 5x year over year (company claims). Founded in 2018 in San Francisco.
The Verdict
Who Should Use Labelbox?
Best For
- AI labs buying expert human data, RL environments and model evaluation
- Enterprises that want software plus a managed expert workforce (Alignerr)
- Teams labeling many modalities in one tool, from images to agent traces
Not Ideal For
- Teams that need self-hosted or open-source labeling (see Label Studio)
- Small teams on a budget: G2 reviewers flag pricing for small teams
- Buyers who want a public price list (the pricing page was offline when we checked)
What's Great
- Ease of use and a clean interface are the top G2 themes
- Data validation and QA tooling
- Software and managed expert labeling from one vendor
- Free tier: 500 LBU a month, up to 30 users and 50 projects
Watch Out For
- Slow performance on large datasets and feature gaps (G2)
- The public pricing page returned 404 in November 2026; plan rates now appear only in the docs
- Focus has shifted toward frontier-lab services, so check the self-serve roadmap
Pricing
View all features & details
Platform
- Annotate: 10+ built-in editors, ontologies, ground-truth import
- Catalog: curation, search, bulk classification, 25+ cloud data sources
- Model: run custom models, import predictions, error analysis
- Foundry: foundation models to generate RLHF preference data
- MCP connection to query progress and quality from an AI assistant
Services and Limits
- Alignerr expert labeling community, 30+ languages
- API: 1,500 requests per minute on Free, 5,000 on Starter, 20,000 on Enterprise
- Ground-truth import: 10,000 on Free, 2M on Starter, 10M on Enterprise
Real-World Usage
Named Customers
- Logos: Shutterstock, Ideogram, Stryker, Intuitive, Warner Bros., Peloton, Pinterest, Walmart, Etsy, Runway, JPL
- Case studies: Meta, Google Cloud (Vertex AI LLM evaluation), Intuitive Surgical
Funding and Reviews
- Series E, more than $100M, October 14, 2026, led by Infinitum Partners (equity plus credit)
- G2: about 4.5 out of 5 from about 48 reviews
How It Compares
| Feature | Labelbox | Encord | Label Studio | Snorkel AI |
|---|---|---|---|---|
| Main focus | Frontier data, RL environments, labeling platform | Physical AI data platform | Open-source labeling and AI evaluation | Expert data-as-a-service |
| Free tier | 500 LBU a month | Not confirmed | Open-source Community Edition | No |
| Open source | No | No | Yes, Apache-2.0 | Old snorkel library only |
| Managed workforce | Alignerr | Not covered here | HumanSignal Data Services | Expert network |
| G2 rating | About 4.5 | About 4.8 | Too few reviews | Too few reviews |
Sources: labelbox.com · Labelbox docs · HumanSignal pricing · snorkel.ai
Is this your tool? Add a badge