What Gamelab.com does

GameLab is an AI training data and benchmarking platform that operates as a dedicated division within Arkadium, Inc., a privately held New York–based game publisher. The platform is built on a wholly-owned portfolio of more than 100 games across 20+ genres, which collectively generate 1.7 billion game plays, 150 billion human decisions, and 200 billion gameplay images per year from 22 million monthly active players. Its core technical premise is that structured, sequential human gameplay data — captured as move-by-move decision trees, full-state records, and outcome-based trajectories — provides AI models with the multi-step reasoning context, consequences, and strategy that static internet-scraped datasets lack.

GameLab sells to frontier AI labs and research institutions through a five-product suite: Human Data (proprietary sequential decision records delivered in JSON/CSV), Bespoke Solutions (custom game and dataset design), Environments (interactive training spaces for RL and agentic evaluation), Benchmarking (the proprietary Cognitive Index Score spanning eight cognitive dimensions), and RL & Training (RLVR and fine-tuning workflow support). The company also operates a public AI Model Leaderboard that ranks frontier models across dynamic game environments. Pricing is quote-based with no publicly disclosed tiers, and distribution is direct-to-enterprise through gamelab.com, with a single direct sales channel described as global in scope.

Named customers include Google DeepMind and Meta AI as core-tier frontier model partners. GameLab is privately held through its parent Arkadium, with no external venture funding disclosed for the division itself, and operates with a small dedicated team (1-10 employees). The parent company's profitable, stable operations provide GameLab with continuous access to live gameplay data, production infrastructure, and an active player ecosystem.

Name

Gamelab.com

Legal name

Arkadium, Inc.

Website

https://gamelab.com

Company type

Private

Operating status

Operating

Headcount range

1–10 employees

Short description

GameLab is an AI training data and benchmarking platform that licenses proprietary human gameplay decision data, custom environments, and cognitive benchmarks to frontier AI labs and research institutions, built on Arkadium's 100+ game portfolio and 22 million monthly players.

Ownership category

akta.pro rank

Product category

AI Training Data & Evaluation Infrastructure

NAICS

Software Publishers (5132), Doll, Toy, and Game Manufacturing (339930), Doll, Toy, and Game Manufacturing (33993)

SIC

Services-Computer Programming, Data Processing, Etc. (7370), Games, Toys & Children's Vehicles (No Dolls & Bicycles) (3944), Services-Computer Programming Services (7371)

akta.pro primary industry

Data Platforms, BI/Reporting & Player Analytics (CDP/Attribution) (MPAEAKAI)

akta.pro secondary industries

Serious Games & Simulation Studios (Training, Defense, Enterprise) (MPAFAAAH), Gaming & Interactive Entertainment Personalization (HDAAAGAF)

Keywords

• AI training data
• LLM benchmarking
• Human decision datasets
• Reinforcement learning environments
• AI reasoning evaluation

Headquarters

HQ city

New York

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations

Revenue model

1. Human Decision Data Licensing: Licensing of structured sequential gameplay data from 22 million monthly players, including full-state game records, move-by-move decision trees, and outcome-based trajectories delivered in JSON/CSV formats
2. Bespoke Dataset Solutions: Custom game environment design and specific gameplay distribution curation for unique research objectives, including edge case scenarios and expert-level play data
3. Benchmarking Services: Cognitive Index Score benchmarking system providing AI model evaluation across eight brain functions, measuring true reasoning capabilities beyond static test sets
4. RL & Training Environments: Training environment access and support for reinforcement learning with verifiable rewards, fine-tuning workflows, and agentic evaluation

Go-to-market motion1 record

Distribution channels1 record

Marketing channels4 records

Core offering

GameLab.com licenses proprietary human decision-making data generated from its owned portfolio of 100+ games (22M monthly players, 1.7B annual plays, 150B human decisions) to frontier AI labs and research institutions. It also builds and operates bespoke AI training environments, an LLM benchmarking platform (Cognitive Index Score / AI Model Leaderboard), and reinforcement learning integrations tailored to enterprise model developers.

Product overview

GameLab is a platform-plus-products architecture within Arkadium, providing AI training data, benchmarking, and interactive game environments for AI reasoning development. The core platform rests on a wholly owned portfolio of 100+ games generating 1.7 billion game plays and 150 billion human decisions per year. Its product suite comprises five integrated offerings: Human Data (proprietary sequential decision records), Bespoke Solutions (custom game and dataset creation), Environments (interactive training spaces), Benchmarking (Cognitive Index Score across eight cognitive dimensions), and RL & Training (end-to-end training workflow support). The AI Model Leaderboard provides a public facing benchmark of frontier model performance, while the Games environment portal gives researchers access to all 100+ game environments as testbeds.

Differentiator

Problem solved

Functional benefit

Brands

• Human Data: Proprietary decision data from 22 million monthly players across 100+ games for AI training.
• Bespoke Solutions
• Environments
• Benchmarking
• RL & Training

Products and services

• Human Data Licensed access to proprietary, large-scale human decision-making datasets (150B+ sequential decisions) captured from GameLab's owned portfolio of 100+ games with 22M monthly players, sold to AI labs for training and evaluating foundation models.
• Bespoke Solutions Custom-built AI training environments, datasets, and evaluation tooling designed and delivered to individual AI lab customers on a contract basis.
• Environments Interactive game-based AI training environments built on GameLab's owned games, provided to AI labs for reinforcement learning, agent training, and model evaluation.
• Benchmarking (Cognitive Index Score) Subscription-based LLM benchmarking platform that scores frontier AI models against real human gameplay performance across cognitive dimensions, providing standardized evaluation of model reasoning.
• RL & Training Reinforcement learning and model training integrations that connect AI labs directly to GameLab's game-based environments and human decision data for post-training and fine-tuning.
• AI Model Leaderboard Public leaderboard that ranks frontier LLMs against real human gameplay performance using the Cognitive Index Score, providing transparent comparison of model reasoning capabilities.

Quantifiable outcome

• 1.7 billion game plays and 150 billion human decisions per year
• +1 more outcomes

Named customers2 records

Segments2 records

Ideal customer profiles1 record

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability7 records

Feature3 records

Partnerships

Three partnerships are on record, tiered core and minor.

• Google DeepMindcoreTechnology or IntegrationFrontier Model Partner using GameLab's cognitive benchmarking and training data services for AI model evaluation and reasoning capability assessment.
• Meta AIcoreTechnology or IntegrationFrontier Model Partner using GameLab's cognitive benchmarking and training data services for AI model evaluation and reasoning capability assessment.
• Good Start LabsminorOthersListed in special thanks section as a research collaborator, indicating informal research partnership or academic collaboration.

Scale indicators7 records

Recent moves6 records

Expansion highlights5 records

Broad incumbents

• Appen: Appen is a long-established provider of training data and data annotation services for AI and machine learning. It is a broad incumbent competing in the same AI data market as GameLab, though across many verticals beyond games and reasoning.
• TELUS International AI Data Solutions: TELUS International (formerly Lionbridge AI) is a large AI data services provider offering annotation, data collection, and evaluation services. It is a broad incumbent operating in the same training-data category as GameLab, with a global services footprint.

Direct peers

• iMerit: iMerit provides high-quality data annotation, labeling, and enrichment services for AI and machine learning applications. It is a comparable AI data services provider targeting similar enterprise and frontier-lab customers.
• Surge AI: Surge AI provides high-quality human-generated training data with a focus on reasoning, evaluation, and RLHF data for frontier model developers. It directly competes with GameLab in supplying structured decision and reasoning data to AI labs.
• Snorkel AI: Snorkel AI provides a data-centric AI platform for programmatic labeling, weak supervision, and dataset creation for model training. It competes with GameLab in the broader AI data infrastructure market for enterprise model developers.
• Labelbox: Labelbox is a data platform for AI development that supports data labeling, evaluation, and benchmarking workflows. It is directly comparable to GameLab's data and benchmarking product layer, serving similar AI developer customers.
• Scale AI: Scale AI is the largest AI training data platform, providing labeled datasets, RLHF data, and evaluation services to frontier AI labs. It is a direct competitor to GameLab's Human Data and RL & Training offerings, addressing the same enterprise customers.

Emerging players

• Encord: Encord provides a data-centric AI platform for data curation, labeling, and model evaluation. It is an emerging player addressing the same AI training data and benchmarking workflows as GameLab, with particular strength in multimodal and agentic use cases.
• Gretel.ai: Gretel.ai is a synthetic data platform that generates privacy-safe training data for AI development. It is an emerging player in the AI training data market, addressing similar customer pain points around data quality, privacy, and contamination.
• Synthesis AI: Synthesis AI produces synthetic training data and simulation environments for computer vision and AI model development. It is an emerging player in the synthetic data and environment space, adjacent to GameLab's game-based environment offering.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks7 records

Key highlights7 records

Customer concentration

Compliance4 records

Revenue estimate

Valuation estimate

Number of profiles

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Gamelab.com