Surge AI
Surge AI is a human-data company that provides large-scale, expert-quality labeled data for training and evaluating frontier AI models. The product surface spans RL environments and agents (rich, complex environments that challenge agentic models), rubrics and verifiers (scoring systems for AI outputs), RLHF (preference and reward data), SFT (foundational skill demonstrations), human evaluation, expert professional domains, internationalization across 70+ languages, multimodal (image, audio, video) data, and off-the-shelf datasets. Surge ships an official Python SDK (surge-python) wrapping the Surge API, with API-key authentication, and exposes the dashboard and API reference at app.surgehq.ai. Public datasets published by Surge include the toxicity dataset (the world's largest social-media toxicity dataset).
APIs
Surge API
Surge's REST API for managing labeling projects, tasks, and results. Endpoints cover projects (list, retrieve, create, download results, save reports in multiple formats), tasks...
Surge Python SDK
Official Python SDK (surge-api on PyPI) wrapping the Surge API. Requires Python 3.10+, MIT-licensed, and last updated May 2026. Configured via surge.api_key or the SURGE_API_KEY...
Surge RL Environments and Agents
Surge's product surface for delivering complex reinforcement-learning environments and agents that challenge and evaluate agentic models.
Surge Rubrics and Verifiers
Scoring rubrics and automated verifiers for grading AI outputs across domains.
Surge RLHF
Preference and reward data for reinforcement learning from human feedback.
Surge SFT
Foundational-skill demonstration data for supervised fine-tuning.
Surge Human Evaluation
Quality assessment of AI outputs by Surge's expert workforce.
Surge Multimodal Data
Image, audio, and video data collection and labeling.
Surge Internationalization
Multilingual data across 70+ languages for localization, translation, and multilingual model evaluation.
Surge Off-The-Shelf Data
Pre-built datasets ready for licensing and download.
Surge Toxicity Dataset
The world's largest open social-media toxicity dataset, published under MIT license.
Collections
Surge AI API
OPENPricing Plans
Rate Limits
FinOps
Surge Ai Finops
FINOPSFeatures
Endpoints for projects, tasks, and blueprints, with API-key authentication.
Official surge-python SDK on PyPI, MIT-licensed, Python 3.10+.
Complex environments that challenge agentic models.
Scoring systems for AI outputs across domains.
Preference, reward, and demonstration data for foundation-model training.
Expert workforce grades AI output quality.
Specialized expertise across finance, law, medicine, and more.
Internationalization coverage spanning more than 70 languages.
Image, audio, and video collection and labeling.
Pre-built datasets available for licensing.
Public releases including the world's largest social-media toxicity dataset.
Use Cases
Preference and reward data for reinforcement learning from human feedback.
Demonstration data for SFT across professional domains.
Benchmark agents in complex RL environments with structured rubrics.
Evaluate model quality across 70+ languages.
Use the toxicity dataset and human evaluation pipelines for trust and safety work.
Integrations
Programmatic integration via the official surge-python SDK.
Standard API-key auth (SURGE_API_KEY env var or programmatic configuration).
Use Surge blueprints as templates for new labeling projects.