Toloka review: the robotics data company for buying data now
Answer in brief: Toloka is the fastest route in this directory to robotics data you can use today. Its off-the-shelf catalog includes 25,000+ hours of egocentric household video, 500+ hours of UMI manipulation data, teleoperation and exocentric sets, with samples within 24 hours. Its self-serve platform runs egocentric collections with no sales cycle and no minimums. Licensing terms and rates are not published.
Directory category: Dataset marketplace · also Egocentric video. Last updated: 2026-11-09. This is independent research from Toloka's public pages and its Hugging Face dataset; we have not licensed or tested Toloka data.
At a glance: 25K+ hours of physical AI data · 216,000+ egocentric clips across 24 countries · samples in 24 hours · self-serve with 200,000+ contributors · open HomER dataset on Hugging Face.
What Toloka does for robotics data
Toloka calls itself the human layer for Physical AI. It supplies data, demonstrations, annotation and evaluation, and it works in three ways: licensing off-the-shelf datasets, running custom data programs, and a self-serve collection platform.
Its physical AI catalog lists four sets. Egocentric household video: 25,000+ hours and 216,000+ clips across 12 activity categories and 24 countries, graded into capture-quality tiers with seven time-aligned annotation layers. UMI manipulation: 500+ hours from a head camera and two hand-held gripper cameras at 1080p30, a 100 Hz IMU, 6-DoF pose and gripper aperture in millimeters. Teleoperation: VR-driven bimanual and single-arm robots with 25 Hz state and actions and wrist and head RGB-D. Exocentric video: third-person capture in studios, homes and commercial spaces. UMI ships as MCAP plus JSON, and video as MP4 plus JSON.
Who it fits
- Teams that need pre-training or imitation-learning data this month, not next quarter.
- Researchers who want to test egocentric collection cheaply before committing a budget.
- Buyers who want samples before any contract.
Strengths: data on the shelf, collection on demand
- Speed: Samples within 24 hours. Toloka says pipeline-ready datasets go live within 48 hours.
- Self-serve: Describe the task and an AI assistant configures the pipeline and selects contributors. Toloka cites 200,000+ contributors in 100+ countries and about 3 hours to a first batch, with no sales cycle and no minimums.
- Public proof: Toloka's team built HomER on its own platform, an open egocentric dataset of 63 videos across 17 household task categories. Toloka says it took 3 hours to collect and cost $50.
- Commercial use: Toloka says its catalog data can be used for commercial training.
Honest drawbacks
- Most of the catalog targets LLM and agent training, not robotics.
- Licensing terms, catalog prices and self-serve rates are not published.
- Self-serve QA is LLM-based. Toloka reports 89.1% accuracy at catching failures, so some bad clips can slip through.
- The HomER sample is small (63 videos), so treat it as a demo rather than a training corpus.
Pricing model
Dataset licensing and custom programs go through sales; the request form offers budget bands starting under $25k. Self-serve collection is pay-as-you-go, but rates are not published.
Best alternatives to Toloka
- Firsthand: richly instrumented custom egocentric capture with a free sample pack.
- Fax Research: licensed existing footage with provenance and authenticity checks.
- Knoxlabs: US in-home egocentric and glove capture, with samples within one business day.
Toloka FAQ
Can I license robotics datasets from Toloka?
Yes. Toloka's catalog lists egocentric household video, UMI manipulation, teleoperation and exocentric video datasets, and it says samples arrive within 24 hours.
What is HomER?
HomER is an open-source egocentric household dataset that Toloka's team built on its self-serve platform. It is on Hugging Face with 63 videos across 17 task categories.
Do I need a sales call to collect egocentric video with Toloka?
No. Toloka's self-serve egocentric video platform has no sales cycle and no minimums.
Sources
- toloka.ai/physical-ai
- toloka.ai/ots-datasets
- toloka.ai/egocentric-video
- huggingface.co/datasets/toloka/HomER
← Back to the robotics data companies directory ·
Named on this page as a provider? Join the network for a linked profile, Visit buttons and featured placement.