High-quality human data for AI training, evaluation, and research
Graviti records 755 monthly visits with -82.8% month-over-month growth and Freemium pricing. Label Studio records 137,199 monthly visits with -16.1% growth and Contact for Pricing, Free pricing.

Real-world training data for Physical AI, robotics and LLMs
| Dimension / Metric | ![]() | ![]() |
|---|---|---|
| Official Domain | graviti.com | labelstud.io |
| Pricing Model | Freemium | Contact for Pricing, Free |
| Monthly Visits | 755 (755) | 137.2K (137,199) |
| MoM Traffic Delta | -3.62K (-82.8%) | -26.29K (-16.1%) |
| Primary Market & Share | 🇮🇳 India (100%) | 🇺🇸 United States (23.9%) |
| Direct Visits / mo | 290 | 47.14K |
| User Rating | 3.8 | 4.7 |
| Supported Platforms | Website | Website |

Real-world training data for Physical AI, robotics and LLMs
Flagship datasets pair a head-mounted camera with multiple fixed cameras recording the same session, synchronized and calibrated so models learn spatial action understanding.
Bimanual teleoperation trajectories, body keypoints, action segments and language instructions are captured and annotated for robot manipulation and world-model research.
For language models, Graviti offers expert-written STEM problems, human preference pairs for RLHF and DPO, agentic coding environments and expert-graded finance, legal and office tasks.
Trained annotators add action segments, hand and body keypoints, narration, preferences and rationales, followed by multi-pass human review, automated checks and contamination screens.

Label any data. Evaluate any AI.
Programmable interfaces and custom templates let one workspace handle images, video, audio, text, PDFs, time series and multi-modal tasks.
Build custom benchmarks and rubrics, run side-by-side model comparisons, and grade generated answers against retrieved sources for RAG evaluation.
Collect human preferences, corrections and rankings to feed reinforcement learning from human feedback and fine-tuning workflows.
Connect observability tools, storage and any model to power AI-assisted pre-labeling, benchmarking and continuous evaluation.
High-quality human data for AI training, evaluation, and research
Free / Open-sourceRun and build with open AI models locally or in the cloud
Free / Open-sourceRun and build with open AI models locally or in the cloud
Free / Open-sourceOpen-source model and dataset hub with one-click online demos
Free / Open-sourceOpen-source model and dataset hub with one-click online demos

SpecialistEnterprise AI infrastructure where software and expert humans work together
SpecialistEnterprise AI infrastructure where software and expert humans work together

Free / Open-sourceSelf-hosted AI workspace for local and cloud models
Free / Open-sourceSelf-hosted AI workspace for local and cloud models

SpecialistThe environments frontier AI trains on, and the agents that do the recurring work
SpecialistThe environments frontier AI trains on, and the agents that do the recurring work

Human-validated training data and evaluation services for frontier AI models

AI data infrastructure for building production-grade models
AI data infrastructure for building production-grade models

SpecialistTrack, evaluate, and ship AI models and agents in one connected environment.
SpecialistTrack, evaluate, and ship AI models and agents in one connected environment.

Free / Open-sourceAgent observability, evaluation, and continuous improvement for production AI
Free / Open-sourceAgent observability, evaluation, and continuous improvement for production AI
SpecialistThe multimodal data layer for computer vision and physical AI teams
SpecialistThe multimodal data layer for computer vision and physical AI teams

Video data infrastructure for physical AI and robot training
Video data infrastructure for physical AI and robot training