Scale AI
Training data, evaluations and AI applications for labs and governments
Overview
Scale AI provides human-labelled training data through its Data Engine, model evaluations and leaderboards through SEAL, and custom AI applications and agents through Scale GenAI Portfolio and Donovan, for AI labs, enterprises and governments. Meta holds a minority stake. Sales-led; no published pricing.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with Scale AI
- Human-labelled and expert training data (Data Engine)
- Model evaluations and SEAL leaderboards
- Custom enterprise AI apps and agents (GenAI Portfolio)
- Defence AI agents (Scale Donovan)
- Global contributor network via Outlier
Scale AI remains one of the largest suppliers of training data and evaluations to frontier AI labs and governments, with SEAL benchmarks widely cited and defence contracts growing. Meta's 2025 investment and the departure of its founder to Meta cost it some lab customers, but the company reports record new business and has a permanent chief executive again. It is a negotiated, enterprise-only service with no self-serve option. Relevant for organisations building or evaluating large models at scale; small teams should look at self-serve labelling tools instead.
What we found
Scale AI provides training data, evaluation and AI applications for companies and governments building artificial intelligence. Its Data Engine supplies human-labelled and expert-written data for training and fine-tuning large models, drawing on a global contributor network run through its Outlier platform, to which Scale says it has paid more than $1 billion. Its SEAL research lab publishes evaluations and leaderboards, including Humanity's Last Exam and SWE-Bench Pro, and introduced 15 new benchmarks in 2025. Scale GenAI Portfolio, formerly the GenAI Platform, builds custom AI applications and agents for enterprises, and Scale Donovan provides agents for defence; the US Department of War awarded Scale two contracts worth nearly $200 million, and it leads Project Thunderforge. Customers named on its site include Meta, Mayo Clinic, Time, BP and the Center for AI Safety, and it says 90% of leading generative-AI model builders use it.
Founded in 2016 and based in San Francisco, Scale had an eventful 2025. In June Meta invested in the company at a valuation of more than $29 billion; press reports put the stake at 49% and non-voting, for about $14.3 billion. Founder Alexandr Wang left to join Meta and remains Scale's chairman. OpenAI said it had been winding down its work with Scale, and in July Scale laid off about 200 employees, 14% of its staff, and 500 contractors. Jason Droege, who led the company as interim chief executive, reported that 2025 was its strongest financial year, with well over $1 billion in new business, a profitable data business and the Middle East as its fastest-growing region. On 10 August 2026 Francis deSouza, formerly president of security products at Google Cloud and chief executive of Illumina, became chief executive.
Scale no longer offers self-serve tools or published prices: every pricing link leads to a demo request, and contracts are negotiated. It suits AI labs, large enterprises and government agencies that need large-scale data and evaluation.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Scale AI's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Scale AI unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Scale AI directly →
Spotted something out of date? Suggest an update →
Alternatives to Scale AI
More in Research & Data
