Kinds of data
Expert and human work
Data that has to be made by people: expert tasks and judgments, narrated recordings of real work, documents their authors own.
- Public price
- $55 to $101
per hour of work
- Demand
- Rising
- Who buys
- Labs and the vendors that staff expert projects
The market rate for expert work, by side
How the rates are computedThe median where three or more figures exist, otherwise the figures we hold. Each side is a different point in the market: what the people who make it are paid, what holders are offered, what catalogues ask, what was paid.
$77
per hour of work, median of 105 figures; middle half $55 to $101
$212
per task, median of 4 figures; middle half $167 to $718
$6.00
per document or page, from 1 figure
$0.08
per task, median of 17 figures; middle half $0.032 to $0.22
$19
per hour of work, median of 9 figures; middle half $13 to $94
$2.58
per record, median of 6 figures; middle half $1.25 to $5.29
$12
per demonstration, from 1 figure
What expert work has sold or been offered for
The whole price board- Median advertised hourly pay for Agentic Tasks & RL Environments roles in AI trainingPrice$50 an hour (live median)
per hour, the median of 45 advertised rates
WhenOct 2026What to knowThe median of 45 distinct advertised hourly rates across 9 platforms, from 64 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for AI Evaluation & Rating (RLHF) roles in AI trainingPrice$60 an hour (live median)
per hour, the median of 94 advertised rates
WhenOct 2026What to knowThe median of 94 distinct advertised hourly rates across 16 platforms, from 151 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Coding & Software Evaluation roles in AI trainingPrice$65 an hour (live median)
per hour, the median of 187 advertised rates
WhenOct 2026What to knowThe median of 187 distinct advertised hourly rates across 13 platforms, from 286 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Data Annotation & Labeling roles in AI trainingPrice$30 an hour (live median)
per hour, the median of 84 advertised rates
WhenOct 2026What to knowThe median of 84 distinct advertised hourly rates across 13 platforms, from 138 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Law, Medicine & Finance roles in AI trainingPrice$70 an hour (live median)
per hour, the median of 684 advertised rates
WhenOct 2026What to knowThe median of 684 distinct advertised hourly rates across 20 platforms, from 1,098 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Math & STEM roles in AI trainingPrice$65 an hour (live median)
per hour, the median of 192 advertised rates
WhenOct 2026What to knowThe median of 192 distinct advertised hourly rates across 12 platforms, from 227 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Multilingual & Localization roles in AI trainingPrice$65 an hour (live median)
per hour, the median of 705 advertised rates
WhenOct 2026What to knowThe median of 705 distinct advertised hourly rates across 13 platforms, from 1,010 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Red-Teaming & AI Safety roles in AI trainingPrice$55 an hour (live median)
per hour, the median of 30 advertised rates
WhenOct 2026What to knowThe median of 30 distinct advertised hourly rates across 8 platforms, from 52 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Research Studies & Surveys roles in AI trainingPrice$100 an hour (live median)
per hour, the median of 17 advertised rates
WhenOct 2026What to knowThe median of 17 distinct advertised hourly rates across 3 platforms, from 34 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Robotics & Physical-AI Data roles in AI trainingPrice$10 an hour (live median)
per hour, the median of 59 advertised rates
WhenOct 2026What to knowThe median of 59 distinct advertised hourly rates across 6 platforms, from 64 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Transcription, Audio & Speech roles in AI trainingPrice$30 an hour (live median)
per hour, the median of 334 advertised rates
WhenOct 2026What to knowThe median of 334 distinct advertised hourly rates across 13 platforms, from 521 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Median advertised hourly pay for Writing & Content roles in AI trainingPrice$55 an hour (live median)
per hour, the median of 20 advertised rates
WhenOct 2026What to knowThe median of 20 distinct advertised hourly rates across 4 platforms, from 37 open roles, recomputed daily. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs pay explorerA catalogue or marketplace listing
- Biologist, PhD at Alignerr (PhD biologists writing and reviewing advanced AI evaluation questions, posted for Labelbox)Price$90-120/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe PhD tier. As on the Chemist, PhD post, the body still carries an older line reading $25 to $100 per hour that disagrees with the $90-120/hr in the header and in the board's own data (lower bound 90, upper bound 120); the header figure is the one the board shows.SourceAlignerr job board, Biologist, PhDThe payer's own page
- Biology Masters at Alignerr (Master's-level biologists evaluating AI answers)Price$35-55/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe Master's tier. About 40 listings carry this title and band, and the 'Biology Subject Matter Expert - AI Content Specialist' posts match it. The Masters/PhDs tier on the same board reads $75-90/hr (alignerr-biology-expert-hourly).SourceAlignerr job board, Biology MastersThe payer's own page
- Chemist, PhD at Alignerr (PhD chemists writing and reviewing advanced AI evaluation questions, posted for Labelbox)Price$90-120/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe PhD tier, the highest chemistry band on the board. The posting's body still carries an older line reading $25 to $100 per hour, which disagrees with the $90-120/hr in its header and in the board's own data (lower bound 90, upper bound 120); the header figure is the one the board shows.SourceAlignerr job board, Chemist, PhDThe payer's own page
- Chemistry Masters at Alignerr (Master's-level chemists evaluating AI answers)Price$35-55/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe Master's tier. About 40 listings on Alignerr's board carry this title and the same band, and the 'Chemistry Subject Matter Expert - AI Content Specialist' posts match it. The Masters/PhDs tier on the same board reads $75-90/hr (alignerr-chemistry-expert-hourly).SourceAlignerr job board, Chemistry MastersThe payer's own page
- CVAT annotation services, one-time project: per annotated object, from its worked example (minimum project $5,000)Price$0.10/object
per annotated object
WhenOct 2026What to knowCVAT's own worked example: 10,000 images at about 10 objects each is 100,000 objects, $10,000 at this rate. It bills per annotated object after a proof of concept and requires a minimum project value of $5,000.SourceCVAT.ai, annotation services pricing explainerA vendor's published price guide
- CVAT annotation services, 6-month prepaid subscription (from $5,000): reduced per-object ratePrice$0.05–0.075 per object
per annotated object (prepaid 6-month subscription)
WhenOct 2026What to knowThe prepaid subscription starts at $5,000 for 6 months, for a buyer who sends data gradually and gives a volume estimate; the same worked example comes to $5,000 to $7,500 for 100,000 objects. The high end of the range is $0.075.SourceCVAT.ai, annotation services pricing explainerA vendor's published price guide
- Label Your Data: bounding box annotation, starting pricePrice$0.02 per object
per object (starting price)
WhenOct 2026What to knowA 'starts at' price that depends on project complexity and volume; the page also offers a cost calculator and a free pilot. The same page's $6 per annotator hour is held as labelyourdata-annotator-hour-bill-rate.SourceLabel Your Data, data annotation pricingA vendor's published price guide
- Label Your Data: keypoint annotation, starting pricePrice$0.015 per object
per object (starting price)
WhenOct 2026What to knowA 'starts as low as' price for keypoints on images; the actual price depends on annotation type, data volume and complexity.SourceLabel Your Data, data annotation pricingA vendor's published price guide
- Label Your Data: NLP annotation, starting price per entityPrice$0.02 per entity
per entity (starting price)
WhenOct 2026What to knowA 'starts as low as' price for NLP tasks, counted per entity to label; the page's calculator also prices by characters or by annotation hours.SourceLabel Your Data, data annotation pricingA vendor's published price guide
- Mercor's average contracted expert rate (homepage)Price$112/hr Average contracted rate
per hour (paid to the worker)
WhenOct 2026What to knowPlatform-stated average of contracted rates, up $1 from $111/hr on 15 October 2026 ($110 on 14 October, $109 on 11 October). The homepage's own data field reads 112.SourceMercor homepageThe payer's own page
- Precise BPO: bounding box annotation, per-task rate range (dedicated offshore team)Price$0.01 – $0.06
per bounding box (rate range)
WhenOct 2026What to knowThe cheapest published image-box range found: a low-wage dedicated team (its hourly model is $4 to $8 an hour). The same table lists polygon or instance at $0.08 – $0.35 and segmentation at $0.10 – $0.45.SourcePrecise BPO, data labeling services pricingA vendor's published price guide
- Roboflow Outsource Labeling: bounding box annotation, starting ratePrice$0.10 per bounding box
per bounding box (starting rate)
WhenOct 2026What to knowRoboflow Outsource Labeling: pay per annotation, with discounted prepaid options; it requires an active Roboflow subscription. Pricing and turnaround vary by project.SourceRoboflow Labeling (outsourced labelling)A vendor's published price guide
- Roboflow Outsource Labeling: classification or keypoint annotation, starting ratePrice$0.05 per classification or keypoint annotation
per classification or keypoint annotation (starting rate)
WhenOct 2026What to knowRoboflow Outsource Labeling: pay per annotation, with discounted prepaid options; it requires an active Roboflow subscription. Pricing and turnaround vary by project.SourceRoboflow Labeling (outsourced labelling)A vendor's published price guide
- Roboflow Outsource Labeling: polygon annotation, starting ratePrice$0.20 per polygon
per polygon (starting rate)
WhenOct 2026What to knowRoboflow Outsource Labeling: pay per annotation, with discounted prepaid options; it requires an active Roboflow subscription. Pricing and turnaround vary by project.SourceRoboflow Labeling (outsourced labelling)A vendor's published price guide
- US-DATA: image bounding box annotation, indicative price per objectPrice1 object from $0.08
per object (indicative 'from' price)
WhenOct 2026What to knowThe page calls its prices indicative and 'not a public offer'; the final cost is set after task analysis and a pilot.SourceUS-DATA, AI data annotation pricingA vendor's published price guide
- US-DATA: named entity recognition (NER) text annotation, indicative price per textPrice1 text from $0.25
per text (indicative 'from' price)
WhenOct 2026What to knowAn indicative 'from' price from the same table that lists text classification at $0.06 a text and dialogue annotation at $1.25 a dialogue; not a public offer.SourceUS-DATA, AI data annotation pricingA vendor's published price guide
- Lawyers (Mercor listing)Price$100-$120 per hour
per hour (paid to the worker)
WhenOct 2026What to knowGeneric lawyer pool listing, US only; the band rose from $70-$110 (read 11 Oct 2026). Paid by Mercor; the lab bill rate is higher. Closed: Mercor shows the posting no longer accepting applications (read 2026-10-17).SourceMercor job boardThe payer's own page
- Chemistry Specialist, Freelance AI Trainer Project (Meridial, global remote contractor)Price$8-to- $65 per hour
per hour (paid to the worker)
WhenOct 2026What to knowOne band for the whole Freelance AI Trainer Project programme: the same $8 to $65 an hour appears on Meridial's Biology Specialist, Analytical, Organic/Inorganic, Synthetic and Radiochemistry posts and on its Yoruba, Hausa and Igbo language posts, with the exact rate set by experience and geographic location. Its Microbiology, Neuroscience and plain Medicine Specialist posts start at $6. The floor is a geography effect, not a skill price.Source - All AI-training work: live median advertised hourly pay across 26 platforms (AITraining.jobs pay report)Pricemedian $65/hr; typical $35-$120/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of 2,493 distinct advertised rates across 26 platforms from 3,725 open roles (1,960 rates, 21 platforms and $60 on 14 Oct). Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026A catalogue or marketplace listing
- Handshake AI: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$90/hr median; typical $55-$150/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates Handshake AI states in its live job feed, first listed in the report's platform table on 15 Oct 2026; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Terac: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$85/hr median; typical $55-$105/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates Terac states in its live job feed; one platform's rate card, not a market rate. Up from $75 (typical $50-$100) on 14 Oct. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Dutch Audio Transcriptionist (AI Training) at AlignerrPrice$10-35/hr
per hour (paid to the worker)
WhenOct 2026What to knowReposted under a new address with the floor cut from $15 to $10 an hour (the old posting now answers 404); native Dutch speakers in the US.SourceAlignerr job board, Dutch Audio TranscriptionistThe payer's own page
- Medical expert work on Mercor, nurses and other clinical professions (standing talent network band)Price$60 to $180 an hour
per hour (paid to the worker)
WhenOct 2026What to knowThe posting closed; its page now shows the standing Mercor talent network for the field, which states the band its listings have paid and names a project's own rate only when it invites someone (was $55 to $65 an hour for registered nurses on the posting). The band spans professions, so it is wider than any one role's rate.SourceMercor, RN Annotators (closed) and its Medical Expert talent networkThe payer's own page
- Physician (MD/DO) work on Mercor, clinical documentation and AI evaluation (standing talent network band)Price$115-$150 per hour
per hour (paid to the worker)
WhenOct 2026What to knowThe posting closed; its page now shows the standing Mercor talent network for the field, which states the band its listings have paid and names a project's own rate only when it invites someone (was $170 an hour on the posting).Source - Babel Audio: median advertised hourly pay across its live rolesPrice$105/hr median; typical $105-$150/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the hourly rates Babel Audio states in its live job feed, measured by AITraining.jobs. First listed among the platforms with enough rates to measure on 14 October 2026. AITraining.jobs has the same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Mindrift (Toloka): median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$100/hr median; typical $100-$105/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates Mindrift (Toloka) states in its live job feed, with the 25th to 75th percentile band $54-$130/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- RemoExperts: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$95/hr median; typical $70-$150/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates RemoExperts states in its live job feed, with the 25th to 75th percentile band $71-$108/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Sovrano AI: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$80/hr median; typical $80-$125/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates Sovrano AI states in its live job feed, with the 25th to 75th percentile band $60-$82/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Data annotation billed by time (Label Your Data)Price$6 per annotator hour
per annotator hour, billed to the buyer
WhenOct 2026What to knowA vendor's own typical hourly bill rate for generalist annotation; undated page read on 2026-10-14.SourceLabel Your Data, Pricing (FAQ)A vendor's published price guide
- Mercor: Consulting Expert (role on the homepage)Price$95/hr
per hour (paid to the expert)
WhenOct 2026What to knowHourly rate Mercor shows for a role among its latest roles on its homepage, which rotates; 70 people recently hired. Closed: The role left Mercor's homepage and is not among its 299 open listings (read 2026-10-17).SourceMercorThe payer's own page
- Mercor: Legal Expert (role on the homepage)Price$90/hr
per hour (paid to the expert)
WhenOct 2026What to knowHourly rate Mercor shows for a role among its latest roles on its homepage, which rotates; 65 people recently hired.SourceMercorThe payer's own page
- Simple annotation on Appen's self-service platformPricerates starting around $0.01–$0.10 per simple annotation
per simple annotation task
WhenOct 2026What to knowSelf-service only; Vendr says managed services are quoted as project totals or retainers and enterprise deals as minimum annual spend against per-task or hourly rates.SourceVendr marketplace, Appen pricing FAQPress report
- Annotation task bought from Appen, typical per-task price across complexityPriceTypical per-task range $0.01–$0.50+ depending on complexity
per task
WhenOct 2026What to knowA procurement marketplace's benchmark, which it says draws on its transaction data; open-ended top. Appen itself publishes custom quotes only. Undated page read on 2026-10-14.Source - Annotation task bought from Scale AI, typical per-task price across complexityPrice$0.05–$1.00+ depending on complexity
per task
WhenOct 2026What to knowSecond column of Vendr's Appen vs. Scale AI table: Scale's typical per-task price runs about two to five times Appen's. Benchmark for commodity annotation tasks, not expert or coding tasks.Source - AI Tutor roles in 13 languages at xAI (Croatian, Finnish, French, Hausa, Italian, Khmer, Korean, Norwegian, Portuguese, Serbian, Spanish, Swedish, Yoruba)Price$35/hour - $45/hour
per hour (paid to the worker)
WhenOct 2026What to knowSame band on every open language tutor posting read on 15 Oct 2026 (French, Spanish, Yoruba quoted); the German, Chinese, Polish, Romanian and Russian postings have closed. US pay; candidates elsewhere get a recruiter estimate. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between.SourcexAI careers (Greenhouse job board), AI Tutor - FrenchThe payer's own page
- BioInformatics Experts at Mercor (writing multi-step bioinformatics data-analysis tasks with validated answers)Price$90-$110 / hr
per hour (paid to the worker)
WhenOct 2026What to knowHourly contract, remote. The work is genomics-style data analysis (public archives such as GEO, SRA and PDB, or the expert's own data) with a reproducing Jupyter notebook per task.SourceMercor job board, BioInformatics ExpertsThe payer's own page
- AfterQuery: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$75/hr median; typical $45-$125/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates AfterQuery states in its live job feed, with the 25th to 75th percentile band $45-$101/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Babel Audio: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$105/hr median; typical $105/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates Babel Audio states in its live job feed, with the 25th to 75th percentile band $105/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Mercor: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$80/hr median; typical $55-$150/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates Mercor states in its live job feed, with the 25th to 75th percentile band $56-$95/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- micro1: median advertised hourly pay across its open AI-training roles (AITraining.jobs)Price$70/hr median; typical $45-$280/hr
per hour (median advertised rate, paid to the worker)
WhenOct 2026What to knowMedian of the rates micro1 states in its live job feed, with the 25th to 75th percentile band $45-$115/hr; one platform's rate card, not a market rate. Same owner as AIData.Broker.SourceAITraining.jobs, The State of AI Training Pay 2026 (by platform)A catalogue or marketplace listing
- Accounting & Finance Expert at AlignerrPrice$60-100/hr
per hour (paid to the worker)
WhenOct 2026What to knowFinance teaching and economist listings on the same board pay $35-70/hr.SourceAlignerr job boardThe payer's own page
- AI Red Team Tester at AlignerrPrice$40-120/hr
per hour (paid to the worker)
WhenOct 2026What to know15 listings at this band on the day read.SourceAlignerr job boardThe payer's own page
- Average pay across Alignerr (Labelbox) expertsPrice$80/hr Average pay
per hour (paid to the worker)
WhenOct 2026What to knowPlatform-stated average; the same page says $1M is paid weekly and 1K+ are hired weekly. Its job board listed 5,704 roles the same day.SourceAlignerr homepageThe payer's own page
- Biology Expert (Masters/PhDs) at AlignerrPrice$75-90/hr
per hour (paid to the worker)
WhenOct 2026What to knowBiology Masters listings on the same board pay $35-55/hr.SourceAlignerr job boardThe payer's own page
- Chemistry Expert (Masters/PhDs) at AlignerrPrice$75-90/hr
per hour (paid to the worker)
WhenOct 2026What to knowChemistry Masters listings on the same board pay $35-55/hr.SourceAlignerr job boardThe payer's own page
- Creative Writing Evaluator at AlignerrPrice$40-120/hr
per hour (paid to the worker)
WhenOct 2026What to knowPrompt Writer listings carry the same band.SourceAlignerr job boardThe payer's own page
- Lean 4 Proof Engineer - Mathematical Formalization (Alignerr)Price$170-200/hr
per hour (paid to the worker)
WhenOct 2026What to knowAbout 130 Lean 4 and formal-verification listings on the board carry this band, the top of Alignerr's STEM pay.SourceAlignerr job boardThe payer's own page
- Legal & Government Policy Analyst (AI Training) at AlignerrPrice$20-40/hr
per hour (paid to the worker)
WhenOct 2026What to knowAlignerr listed only 16 legal roles, nearly all at this band; no attorney-level legal role was open.SourceAlignerr job boardThe payer's own page
- Mathematics Expert (Masters/PhDs) at AlignerrPrice$75-90/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe chemistry and biology Masters/PhD expert listings carry the same $75-90/hr band.SourceAlignerr job boardThe payer's own page
- Machine Learning Evaluation Specialist at Alignerr (designing ML challenges that stump frontier models)Price$200-400/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe highest band on Alignerr's board (7 listings).SourceAlignerr job boardThe payer's own page
- Oncology Clinical Researcher at AlignerrPrice$40-80/hr
per hour (paid to the worker)
WhenOct 2026What to knowThe commonest band among Alignerr's ~800 health and clinical listings; no physician-specific role was listed.SourceAlignerr job boardThe payer's own page
- Physicist, PhD at AlignerrPrice$90-120/hr
per hour (paid to the worker)
WhenOct 2026What to knowMost of Alignerr's 183 physics listings pay less ($35-55/hr for Masters, $40-80/hr for applied physics).SourceAlignerr job boardThe payer's own page
- Software Developer paid to build their own project with an AI coding assistant (Alignerr)Price$60-120/hr
per hour (paid to the worker)
WhenOct 2026What to knowListed in 25 cities at the same band; the developer keeps the code, and the sessions are the data.SourceAlignerr job boardThe payer's own page
- Voice Actor - German Speakers at AlignerrPrice$250-280/hr
per hour (paid to the worker)
WhenOct 2026What to knowGerman voice actors were the best-paid voice roles; Italian and French voice actors listed at $170-200/hr, UK and Australian English at $120-200/hr, Indian English and Mexican Spanish at $40-80/hr.SourceAlignerr job boardThe payer's own page
- Alignerr (Labelbox) weekly payouts to expertsPrice$1M Paid weekly
per week, all experts
WhenOct 2026What to knowAbout $52M a year if sustained (x52); far below Handshake's ~$60M a month.SourceAlignerr homepageThe payer's own page
- Software engineering, data science and ML roles on DataAnnotationPrice$40 – $150+ / hr
per hour (paid to the worker)
WhenOct 2026What to knowPosted pay band for software engineering, data science and ML roles on the platform's own site; the top is open-ended.SourceDataAnnotationThe payer's own page
- Finance, legal, medical, cybersecurity, engineering and STEM roles on DataAnnotationPrice$40 – $125+ / hr; generalists $25 – $50 / hr; language specialists $25 – $40 / hr
per hour (paid to the worker)
WhenOct 2026What to knowSame page sets generalist work at $25 to $50 and language work at $25 to $40, so credentialed fields earn roughly 1.6 to 3 times generalist pay.SourceDataAnnotationThe payer's own page
- Clinical Annotator III: GSA schedule ceiling rate billed to US government buyers (FedPoint Systems)Price$124.84 per hour
per hour, ceiling rate billed to the buyer
WhenOct 2026What to knowA published contract ceiling (MAS contract 47QTCA23D00AR, Bachelors, 4+ years, customer facility), not a paid invoice; a rare public bill rate for annotation labour.SourceGSA CALC+ ceiling-rate API (Multiple Award Schedule)A catalogue or marketplace listing
- Data Labeler III (TS/SCI): GSA schedule ceiling rate billed to US government buyers (Chenega Technology & Modernized Solutions)Price$72.98 per hour
per hour, ceiling rate billed to the buyer
WhenOct 2026What to knowA published contract ceiling (MAS contract 47QTCA26D009E, Bachelors, 5+ years, customer facility), not a paid invoice; a rare public bill rate for annotation labour.SourceGSA CALC+ ceiling-rate API (Multiple Award Schedule)A catalogue or marketplace listing
- Data Labeler, Intermediate: GSA schedule ceiling rate billed to US government buyers (Chenega Systems)Price$94.33 per hour
per hour, ceiling rate billed to the buyer
WhenOct 2026What to knowA published contract ceiling (MAS contract 47QTCA24D0075, Bachelors, 5+ years, customer facility), not a paid invoice; a rare public bill rate for annotation labour.SourceGSA CALC+ ceiling-rate API (Multiple Award Schedule)A catalogue or marketplace listing
- Medical Annotator III: GSA schedule ceiling rate billed to US government buyers (FedPoint Systems)Price$152.72 per hour
per hour, ceiling rate billed to the buyer
WhenOct 2026What to knowA published contract ceiling (MAS contract 47QTCA23D00AR, Bachelors, 4+ years, customer facility), not a paid invoice; a rare public bill rate for annotation labour.SourceGSA CALC+ ceiling-rate API (Multiple Award Schedule)A catalogue or marketplace listing
- Cumulative payouts to Handshake AI fellowsPrice$500M+ payouts made
cumulative total, all fellows
WhenOct 2026What to knowTotal paid to 188k+ fellows since Handshake AI launched in January 2025; a payout total, not a rate.SourceHandshake AI, fellowship homepageThe payer's own page
- Handshake AI data business gross marginPricegross margins around ~30%
share of gross revenue kept after paying experts
WhenOct 2026What to knowResearch-firm figure; the same page puts AI gross revenue near $1B (April 2026) and net after payouts near $300M, consistent with about 30% kept.SourceSacra, Handshake company pagePress report
- Handshake AI experts in mathematics, physics and computer sciencePrice$100–125 per hour
per hour (paid to the expert)
WhenOct 2026What to knowResearch-firm estimate, not a Handshake rate card; Handshake job posts seen the same day ranged from $75 to $100 (chemistry) to $110 to $140 (medicine).SourceSacraPress report
- AI Evaluation Specialist (Handshake AI fellowship, trending opportunity)PriceUp to $17/hr
per hour (paid to the worker)
WhenOct 2026What to knowA ceiling, not a typical rate. The same page lists Game Developer/Designer and PCB Tool Specialist at up to $125/hr and an AI Evaluation Specialist at up to $17/hr.SourceHandshake AI, fellowship homepageThe payer's own page
- FP&A Analyst (Handshake AI fellowship, trending opportunity)PriceUp to $80/hr
per hour (paid to the worker)
WhenOct 2026What to knowA ceiling, not a typical rate. The same page lists Game Developer/Designer and PCB Tool Specialist at up to $125/hr and an AI Evaluation Specialist at up to $17/hr.SourceHandshake AI, fellowship homepageThe payer's own page
- Philosophy Expert (Handshake AI fellowship, trending opportunity)PriceUp to $120/hr
per hour (paid to the worker)
WhenOct 2026What to knowA ceiling, not a typical rate. The same page lists Game Developer/Designer and PCB Tool Specialist at up to $125/hr and an AI Evaluation Specialist at up to $17/hr.SourceHandshake AI, fellowship homepageThe payer's own page
- Software Engineer (Handshake AI fellowship, trending opportunity)PriceUp to $120/hr
per hour (paid to the worker)
WhenOct 2026What to knowA ceiling, not a typical rate. The same page lists Game Developer/Designer and PCB Tool Specialist at up to $125/hr and an AI Evaluation Specialist at up to $17/hr.SourceHandshake AI, fellowship homepageThe payer's own page
- Applied Physics Benchmark Specialist role on Mercor's live boardPrice$61-$77/hr
per hour (paid to the expert)
WhenOct 2026What to knowA benchmark-writing role, priced well below the data science role on the same page. Closed: The role left Mercor's homepage and is not among its 299 open listings (read 2026-10-17).SourceMercorThe payer's own page
- Chemists, explosives and nuclear security specialists writing safety tasks on MercorPrice$65 - $75 / task
per task (paid to the expert)
WhenOct 2026What to knowAbout a dozen hazardous-domain roles (toxicological chemist, blasting engineer, nonproliferation analyst, nuclear engineer) all carried the same per-task band; the generalist role on the same page paid $50 - $60 / hour.SourceMercor job boardThe payer's own page
- Share of Mercor's gross revenue paid to contractorsPrice60-70% of top-line revenue
share of gross revenue
WhenOct 2026What to knowImplies the lab pays about $1.43 to $1.67 for each $1 an expert receives (1 / 0.70 to 1 / 0.60).SourceSacra, Mercor company pagePress report
- Mercor daily payouts to expertsPrice$4M+ Daily payouts
per day, all experts
WhenOct 2026What to knowUp from more than $1.5M a day in October 2025 and more than $2M a day in mid-2026; $4M a day is about $1.46B a year (x365).SourceMercor homepageThe payer's own page
- Data Science Expert role on Mercor's live boardPrice$100-$150/hr
per hour (paid to the expert)
WhenOct 2026What to knowLive listing on Mercor's homepage the day it was read; the same page shows a $109/hr average contracted rate, and listings change daily. Closed: The role left Mercor's homepage and is not among its 299 open listings (read 2026-10-17).SourceMercorThe payer's own page
- Inorganic Chemistry & Synthesis Expert (PhD-level evaluation of materials synthesis problems) at MercorPrice$84 per hour
per hour (paid to the worker)
WhenOct 2026What to knowA single fixed hourly rate, not a band; hourly contract, remote. Mercor pays the expert; the lab's bill rate is higher.SourceMercor job board, Inorganic Chemistry & Synthesis ExpertThe payer's own page
- Mindrift (Toloka) AI trainers, domain expertsPrice$60-100+/hr
per hour (paid to the worker)
WhenOct 2026What to knowOpen-ended top for experts; the same page caps monthly earnings by project at $5,400 (coding) to $8,000 (consulting).SourceMindrift homepageThe payer's own page
- Mindrift (Toloka) AI trainers, entry levelPrice$15-30 per hour
per hour (paid to the worker)
WhenOct 2026What to knowOpen-ended top for experts; the same page caps monthly earnings by project at $5,400 (coding) to $8,000 (consulting).SourceMindrift homepageThe payer's own page
- A software expert recording and narrating real work on screen (Mercor listing)PriceUp to $65 an hour
per hour, paid to the person
WhenOct 2026What to knowWhat the person is paid, not what the lab pays. Vendors pass roughly 60 to 70% of the lab's price to the person.SourceMercorThe payer's own page
- OpenTrain AI managed AI-training teamsPriceCost-plus 20% management fee
fee on top of trainer cost
WhenOct 2026What to knowManaged recruiting, vetting, daily management and QA for 20% over cost: well below the 30-40% gross margins reported for Mercor, Handshake and micro1.SourceOpenTrain AI, PricingA vendor's published price guide
- OpenTrain AI self-service fee on payments to AI trainers and annotatorsPrice10% marketplace fee plus $9.95 per contract
fee on payments to talent, plus a per-contract charge
WhenOct 2026What to knowA published buyer-side markup: 10% on top of the trainer's rate (low/high are the $9.95 initiation fee per contract).SourceOpenTrain AI, PricingA vendor's published price guide
- Cumulative payouts to Outlier (Scale AI) contributorsPrice$500M+ Paid to experts
cumulative total, all contributors
WhenOct 2026What to knowA payout total, not a rate; Outlier publishes no hourly schedule on its homepage.SourceOutlier homepageThe payer's own page
- Prolific participant and tasker pay, minimumPrice$8.00 per hour
per hour (paid to the worker)
WhenOct 2026What to knowProlific recommends higher pay for AI annotation, coding or advanced STEM tasks.SourceProlific, Pricing FAQThe payer's own page
- Prolific participant and tasker pay, recommendedPrice$12.00 per hour
per hour (paid to the worker)
WhenOct 2026What to knowProlific recommends higher pay for AI annotation, coding or advanced STEM tasks.SourceProlific, Pricing FAQThe payer's own page
- Prolific platform fee on participant and AI-tasker rewards (academic or non-profit customers)Price33.3% platform fee
fee on top of worker pay
WhenOct 2026What to knowA published markup: the buyer pays the worker's reward plus 33.3%. Managed services and AI-tasker programmes are quoted privately.SourceProlific, Pricing FAQA vendor's published price guide
- Prolific platform fee on participant and AI-tasker rewards (corporate customers)Price42.8% platform fee
fee on top of worker pay
WhenOct 2026What to knowA published markup: the buyer pays the worker's reward plus 42.8%. Managed services and AI-tasker programmes are quoted privately.SourceProlific, Pricing FAQA vendor's published price guide
- Surge AI expert contractor (award-winning author or journalist tier)Price$200-400 / hour, contractor
per hour (paid to the worker)
WhenOct 2026What to knowThe only rate on Surge's own workforce page, shown above an award-winning author journalist profile; other profiles (constitutional lawyer, investment banker, professor) carry no rate.SourceSurge AI, Workforce pageThe payer's own page
- Oncologist (Mercor listing)Price$350-$450 per hour
per hour (paid to the worker)
WhenOct 2026What to knowThe highest hourly band on Mercor's open board the day it was read. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- A developer's own AI coding session, shared as data (Mercor)Price$200 per task
per task (paid to the worker)
WhenOct 2026What to knowPays for existing work sessions, not new labour: the closest a vendor listing comes to buying data a person already has. Closed: Mercor shows the posting no longer accepting applications (read 2026-10-17).SourceMercor job boardThe payer's own page
- Database Systems Evaluation Expert, per task (Mercor)Price$2.2K per task
per task (paid to the worker)
WhenOct 2026What to knowThe highest per-task price on Mercor's board the day read.SourceMercor job boardThe payer's own page
- Research Advisor, Human Frontier Collective (Scale AI): senior finance, legal and medical expertsPrice$300/hr
per hour (paid to the worker)
WhenOct 2026What to knowPart-time contract for partners, senior counsel, board-certified MDs and bulge-bracket finance seniors; the role also represents Scale with clients. The HFC Fellow roles on the same board state no rate.SourceScale AI careers (Greenhouse), Research Advisor, Human Frontier Collective (US)The payer's own page
- AI Tutor, Video at xAIPrice$40/hour - $75/hour
per hour (paid to the worker)
WhenOct 2026What to knowUS pay; a weekend variant of the role carries the same band. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between.SourcexAI careers (Greenhouse job board)The payer's own page
- AI Red Teamer (LLM generalist, Remote) at Handshake AIPrice$32 - $95 per hour
per hour (paid to the worker)
WhenOct 2026What to knowW-2 hourly through TCWGlobal, 40 hours a week; a Seattle variant carries the same band. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- AI Tutor, Design at xAIPrice$35/hour - $75/hour
per hour (paid to the worker)
WhenOct 2026What to knowUS pay; first published 3 September 2026. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between.SourcexAI careers (Greenhouse job board)The payer's own page
- AI Tutor, Humanities at xAIPrice$35/hour - $75/hour
per hour (paid to the worker)
WhenOct 2026What to knowUS pay; first published 3 September 2026. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between.SourcexAI careers (Greenhouse job board)The payer's own page
- Lean Engineer, formal mathematics (Lean 4, Mathlib) (Mercor)Price$90-$110 per hour
per hour (paid to the worker)
WhenSep 2026What to knowCompare Alignerr's Lean 4 listings at $170-200/hr the same day. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- AI Tutor, Legal and Compliance at xAIPrice$50/hour - $100/hour
per hour (paid to the worker)
WhenSep 2026What to knowThe highest band among xAI's tutor postings; US pay. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between.SourcexAI careers (Greenhouse job board)The payer's own page
- AI Tutor, Software Engineering at xAIPrice$40/hour - $100/hour
per hour (paid to the worker)
WhenSep 2026What to knowUS pay; xAI also lists an Expert Team Lead, SWE at $104,000 to $170,400 a year. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between. Closed: The posting returns 404 and xAI's board has no software-engineering tutor role (read 2026-10-17).SourcexAI careers (Greenhouse job board)The payer's own page
- AI Tutor, Writing at xAIPrice$35/hour - $75/hour
per hour (paid to the worker)
WhenSep 2026What to knowUS pay; the posting says the rate depends on the subspecialty applied for. xAI hires tutors directly, so this is a lab paying the worker with no vendor in between.SourcexAI careers (Greenhouse job board)The payer's own page
- Condensed Matter Physicist (PhD) (Mercor)Price$80-$110 per hour
per hour (paid to the worker)
WhenSep 2026What to knowTen physics PhD subfield listings opened on 25 September 2026 all carry $80-$110. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Biotechnology Expert talent network at Mercor (bioprocess and therapeutic judgment for AI model evaluation)Price$110 to $125 an hour
per hour (paid to the expert, past contract rates)
WhenSep 2026What to knowA talent-network post: the band is what past contracts in the field have paid, set per project. The highest advertised life-sciences band found on Mercor this run.SourceMercor job board, Biotechnology ExpertThe payer's own page
- Chemistry Expert talent network at Mercor (graduate-level chemistry judgment for AI model evaluation)Price$65 to $105 an hour
per hour (paid to the expert, past project rates)
WhenSep 2026What to knowA talent-network post: the band is what past projects have paid, set per project, not a standing offer. Mercor's Biology Expert post uses the same wording at $70 to $105 an hour.SourceMercor job board, Chemistry ExpertThe payer's own page
- Audiobook QA Expert, English (US) (Mercor)Price$20-$25 per hour
per hour (paid to the worker)
WhenSep 2026What to knowFrench, German, Italian, Dutch, Spanish and Portuguese audiobook QA list at $15-$20. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Equity Research Expert, single-name analyst publications (Mercor)Price$150-$250 per hour
per hour (paid to the worker)
WhenSep 2026What to knowCentral bank and company-analysis expert listings carry the same band. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Application user recording generalist software use (Photoshop on Windows) (Mercor)Price$30-$40 per hour
per hour (paid to the worker)
WhenSep 2026What to knowMercor's computer-use series: generalist apps $30-$40, STEM apps $45-$55, code tools $55-$65, office professionals $60-$70. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Molecular Biology Experts, US (Mercor)Price$70-$105 per hour
per hour (paid to the worker)
WhenSep 2026What to knowA UK variant lists the same band. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- AI Red Teamer, CBRNE (Remote) at Handshake AIPrice$65 - $158 per hour
per hour (paid to the worker)
WhenSep 2026What to knowContract, US remote; chemical, biological, radiological, nuclear and explosives red-teaming. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- AI Red Teamer, Cybersecurity (Remote) at Handshake AIPrice$65 - $125 per hour
per hour (paid to the worker)
WhenSep 2026What to knowContract, US remote. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- AI Model Policy Trainer, Image Evaluation (Seattle onsite) at Handshake AIPrice$36 - $72 per hour
per hour (paid to the worker)
WhenSep 2026What to knowOnsite in Seattle. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- AI Safety Policy Evaluator, Violence and Threats (Seattle onsite) at Handshake AIPrice$55 - $120 per hour
per hour (paid to the worker)
WhenSep 2026What to knowContent-policy evaluation for frontier labs; needs trust-and-safety, threat assessment or similar depth. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- Google's annual contract with Surge AIPricemore than $100 million per year
per year
WhenSep 2026What to knowForbes (September 2025) via Contrary.Source - Meta's 2024 spend with Surge AIPrice$150 million in 2024
per year
WhenSep 2026What to knowContrary attributes the figure to The Information. One lab's annual human-data bill with one vendor.Source - Surge AI's prices relative to competing human-data vendorsPrice50% to ten times more than competitors
relative to competitors' prices
WhenSep 2026What to knowForbes (September 2025) via Contrary; Surge's DataAnnotation pays $25-$50/hr generalists and $40-$150+/hr specialists, so the spread there is widest.SourceContrary Research, Surge AI memoPress report
- Drug Discovery Scientist at Mercor (building drug R&D evaluation problems for frontier AI models)Price$60-$90 / hr
per hour (paid to the worker)
WhenAug 2026What to knowHourly contract, remote. Life-sciences work next to biology: experts write research problems wrapped around messy evidence, plus the data room and the grading document.SourceMercor job board, Drug Discovery ScientistThe payer's own page
- AI Model Policy Trainer, Generalist (Seattle) at Handshake AIPrice$30 - $85 per hour
per hour (paid to the worker)
WhenAug 2026What to knowPlacement within the range depends on experience. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- Spanish Tutors (Mercor)Price$60-$65 per hour
per hour (paid to the worker)
WhenAug 2026What to knowA language role priced near expert rates. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- AI Model Policy Trainer, Mental Health (Seattle onsite) at Handshake AIPrice$45 - $65 per hour
per hour (paid to the worker)
WhenAug 2026What to knowOnsite in Seattle, Monday to Friday. Paid by Handshake AI, which sells the work to frontier labs.SourceHandshake careers (Ashby job board)The payer's own page
- Clinicians (Mercor listing)Price$120-$150 per hour
per hour (paid to the worker)
WhenAug 2026What to knowGeneric clinician pool listing. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- micro1 net annual run rate after paying experts, on a $500M gross run ratePricenet annual run rate between $150 million and $200 million
per year, kept by micro1
WhenAug 2026What to know$150M-$200M of $500M is 30-40% kept; the article's own wording ('retains roughly 60% to 70%') contradicts its numbers, so read it as 60-70% paid out. Single unnamed source.SourceTechCrunchPress report
- micro1 gross margin on off-the-shelf datasets resold to several customersPriceas high as 80% to 90%
gross margin on resold datasets
WhenAug 2026What to knowSelling the same dataset more than once more than doubles the 30-40% spread on made-to-order expert work.SourceTechCrunchPress report
- Biochemistry & Life Sciences Domain Expert at Mercor, embedded with a leading AI lab's GenAI team (USA only)Price$65-$105 / hr
per hour (paid to the worker)
WhenAug 2026What to knowA full-time position (Bay Area, CA) advertised at an hourly rate: reviews life-sciences tasks, writes instruction specs and golden solutions and builds benchmarks. Mercor's Pharmaceutical R&D ($75-$115) and Materials Science ($70-$110) domain-expert posts are the same programme at a different band and are not repeated.SourceMercor job board, Biochemistry & Life Sciences Domain ExpertThe payer's own page
- Agent Engineer, per task (Mercor)Price$100-$500 per task
per task (paid to the worker)
WhenAug 2026What to knowAgentic task building, paid per accepted task.SourceMercor job boardThe payer's own page
- Investment Banking & M&A finance domain expert (Mercor)Price$100-$150 per hour
per hour (paid to the worker)
WhenAug 2026What to knowPrivate Equity and Venture Capital domain experts list at $110-$150 per hour. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Partner & General Counsel, legal domain expert (Mercor)Price$100-$150 per hour
per hour (paid to the worker)
WhenAug 2026What to knowSenior Counsel listings pay $85-$120 per hour. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Data annotation billed by the annotator hour, all regions and skill levelsPrice$6 to $60 per annotator hour
per annotator hour, billed to the buyer
WhenAug 2026What to knowWhat a buyer pays the vendor per hour, not what the annotator earns; the spread depends on region and skill. The same table puts medical expert review at $30 to $150+ an hour.SourceHabileData blog, How Much Does Data Annotation Cost? (updated 5 Aug 2026)A vendor's published price guide
- Expert RLHF task such as grading code or medical answers, bought from an annotation vendorPrice$20 to $100+ each
per expert task
WhenAug 2026What to knowAn annotation vendor's own price guide, which it calls indicative and aggregated from public pricing pages and quotes; the top is open-ended, and the same page's FAQ says expert evaluations for code, medical or legal content can exceed $50 per task.SourceHabileData blog, How Much Does Data Annotation Cost? (updated 5 Aug 2026)A vendor's published price guide
- Generalist RLHF preference ranking (one pairwise comparison), bought from an annotation vendorPrice$0.50 – $8.00 per comparison (generalist)
per comparison
WhenAug 2026What to knowVendor price guide, indicative; the page's own worked example budgets an ongoing RLHF pipeline at $4 per comparison.SourceHabileData blog, How Much Does Data Annotation Cost? (updated 5 Aug 2026)A vendor's published price guide
- Professional documents licensed by the person who owns them (Handshake AI)Price$6 per accepted page, up to $30,000
per page
WhenAug 2026What to knowOnly documents the person owns or may share. Work written for an employer or a client usually belongs to them, not to its author.SourceBusiness Insider, via AOLPress report
- AI Safety Expert, English and Norwegian (Mercor)Price$48-$62 per hour
per hour (paid to the worker)
WhenJul 2026What to knowDanish, Finnish, Swedish and Dutch carry the same band. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- AI Safety Expert, English and Vietnamese (Mercor)Price$17-$25 per hour
per hour (paid to the worker)
WhenJul 2026What to knowSame work as the Norwegian listing at about a third of the pay; Indonesian and Malay also list at $17-$25, Thai at $24-$35. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Mercor's average pay to about 30,000 contractors (H1 2026)Pricean average rate of $105 per hour
per hour (paid to the worker)
WhenJul 2026What to knowFrom financial documents reported by The Information; the same report puts H1 2026 gross revenue at $614M with over 90% from foundation model companies.SourceCrypto Briefing (citing The Information)Press report
- Mercor gross margin on AI-lab billings, 2025Price27% (2025)
share of gross revenue kept after paying experts
WhenJul 2026What to knowFull-year 2025. At 27% kept, a lab paid about $1.37 for each $1 Mercor paid an expert (1 / 0.73).SourceBigGo Finance (citing The Information)Press report
- Mercor gross margin on AI-lab billings, Q2 2026Price33% (Q2 2026)
share of gross revenue kept after paying experts
WhenJul 2026What to knowAt 33% kept, about $1.49 billed for each $1 paid to an expert (1 / 0.67); the same report says contractors take roughly two-thirds of revenue.SourceBigGo Finance (citing The Information)Press report
- Mercor gross margin on AI-lab billings, 2027 targetPrice46% (2027 target)
share of gross revenue kept after paying experts
WhenJul 2026What to knowA company projection, not a result: the vendor plans to widen its spread on the same work.SourceBigGo Finance (citing The Information)Press report
- Mercor gross revenue in the first half of 2026, and the share of it from AI foundation model companiesPrice$614 million, more than 90% from AI foundation model companies
gross revenue, first half of 2026 (all customers)
WhenJul 2026What to knowAggregate revenue across all of Mercor's customers, reported from financial documents The Information saw; it shows how concentrated the buying side of expert data is. - AI Safety Red Teamer (Mercor)Price$70-$84 per hour
per hour (paid to the worker)
WhenJul 2026What to knowHazardous-domain safety roles (chemistry, nuclear, explosives) are paid per task instead, at $65-$75. Paid by Mercor; the lab's bill rate is higher. Closed: Mercor shows the posting no longer accepting applications (read 2026-10-17).SourceMercor job boardThe payer's own page
- Share of every dollar kept by data vendors that run delivery from lower-cost geographiesPrice70-80%+
share of every dollar kept
WhenJul 2026What to knowA practitioner's essay. The same essay says a frontier-lab contact works directly with seven data vendors, all tasked with the same type of dataset; that nothing renews by default; and that headline run rates are usually the best month multiplied by 12. - Handshake AI's largest frontier AI lab customer accountsPrice$100M+ run rate per account
per account per year
WhenJul 2026What to knowWhat a single frontier lab pays Handshake a year for human data, in the vendor's own hiring copy; labs are not named.SourceHandshake careers, Senior Engagement Manager postingThe payer's own page
- Share of Mercor's topline revenue paid out to its contractorsPrice60% to 70%
share of topline revenue paid out to contractors
WhenJul 2026What to knowThe mirror of the margin: Mercor keeps 30% to 40% of what a lab pays, so a lab pays roughly 1.4 to 1.7 times the expert's rate.Source - Mercor's advertised headline rate (early June 2026)Price$141/hr in early June 2026
per hour (paid to the worker)
WhenJun 2026What to knowThird-party tracker of the homepage figure: $99/hr in April 2026, $141/hr early June, relabeled and reset to $80/hr late June, $108-$122/hr since.SourceUsagePricing, Mercor pricing blueprintPress report
- Writing expert role on MercorPrice$75 to $150 per hour
per hour (paid to the expert)
WhenMay 2026What to knowThe writer read the rate off Mercor's own site; one listing, not a field average.SourceMoneywise (on Yahoo Finance)Press report
- RLHF preference label for reward modelling, crowdsourcedPrice$0.50 to $2 per ranking
per ranking
WhenApr 2026What to knowAn annotation platform's published cost guide, not a quote from a named vendor; the same sentence prices 50,000 labels at $30,000 to $150,000 in total.SourceTaskMonk blog, Top RLHF Platforms in 2026 (updated 28 Apr 2026)A vendor's published price guide
- RLHF preference label for reward modelling, by domain expertsPrice$5 to $10 per ranking
per ranking
WhenApr 2026What to knowFive to ten times the crowdsourced price in the same sentence: the expert premium on the same unit of work.SourceTaskMonk blog, Top RLHF Platforms in 2026 (updated 28 Apr 2026)A vendor's published price guide
- Human-written SFT demonstration dataset of 10,000 demonstrationsPrice$50,000 to $300,000
per dataset of 10,000 demonstrations
WhenApr 2026What to knowPlatform's cost guide; demonstrations cost more than rankings because annotators write full responses.SourceTaskMonk blog, Top RLHF Platforms in 2026 (updated 28 Apr 2026)A vendor's published price guide
- Nuanced RLHF evaluation requiring domain expertise (one comparison set)Price$20-100+ per comparison set
per comparison set
WhenApr 2026What to knowAn AI services firm's published cost analysis; open-ended top, and ten to twenty times its simple-comparison price on the same page.SourceAI Superior, LLM Data Labeling Cost: 2026 Industry Analysis (17 Apr 2026)A vendor's published price guide
- Simple RLHF preference judgment on a straightforward query (one comparison)Price$2-5 per comparison
per comparison
WhenApr 2026What to knowHigher than NextWealth's $0.10 to $0.50 and TaskMonk's $0.50 to $2 for crowd comparisons: guides disagree by an order of magnitude on the same unit.SourceAI Superior, LLM Data Labeling Cost: 2026 Industry Analysis (17 Apr 2026)A vendor's published price guide
- Enterprise LLM alignment (RLHF) programme, totalPrice$50,000 to several hundred thousand dollars
per programme
WhenApr 2026What to knowEnterprise programmes, far below frontier-lab contracts of $100M+ a year.SourceNextWealth blog, RLHF for Enterprise LLMsA vendor's published price guide
- Domain-expert RLHF annotation (medical, legal, financial)Price$1.50-$8.00+ per task
per task
WhenApr 2026What to knowOpen-ended top; 15 to 16 times the general comparison price at the bottom of each range.SourceNextWealth blog, RLHF for Enterprise LLMsA vendor's published price guide
- General RLHF preference comparison (one pair judged)Price$0.10-$0.50 per comparison
per comparison
WhenApr 2026What to knowA services vendor's published range for buyers; general-domain raters, not experts.SourceNextWealth blog, RLHF for Enterprise LLMsA vendor's published price guide
- Voice Actor for CX agent voice cloning, USA (Mercor)Price$50-$150 per hour
per hour (paid to the worker)
WhenMar 2026What to knowMost of the 15 voice-cloning listings (accents from Peru to Northern UK) pay $50-$100. Paid by Mercor; the lab's bill rate is higher.SourceMercor job boardThe payer's own page
- Chemist Talent Network at Mercor (open application for future AI-training contracts in chemistry)Price$60-$80 / hr
per hour (paid to the worker)
WhenFeb 2026What to knowPart-time, remote. Mercor's Biologist Talent Network and Physical Scientist Talent Network posts show the same $60-$80 / hr band, so it is one programme rate (filed here, not repeated for biology).SourceMercor job board, Chemist Talent NetworkThe payer's own page
- Effective pay for a generalist tasker on some Outlier (Scale AI) projectsPricearound $20 an hour, down from $50
per hour (paid to the worker)
WhenDec 2025What to knowOne tasker's account after Meta's Scale AI deal; shows generalist rates falling while expert rates held.SourceBusiness Insider (on AOL)Press report
- Handshake AI monthly payouts to its experts, against its gross run ratePrice~$60M a month to over 30,000 experts; ~$1B run rate
per month, all experts
WhenNov 2025What to knowThe vendor's own figures: $60M a month is $720M a year (x12), about 72% of a ~$1B run rate, so roughly 28% stays with Handshake; about $2,000 per expert per month ($60M / 30,000). Repeated in a September 2026 red-teamer posting.SourceHandshake careers, Engagement Manager postingThe payer's own page
- Generalist data work on Mercor's Meta project Nova, after project Musen endedPrice$16 an hour
per hour (paid to the worker)
WhenNov 2025What to knowMusen paid $5 more ($21 an hour) and had more than 5,000 workers at peak; Mercor called the reporting inaccurate. The low end of Mercor's pay, for a Meta generalist project.SourceBusiness Insider (on AOL)Press report
- Mercor's top industry-expert rate (October 2025)Priceup to $200 an hour
per hour (paid to the worker)
WhenOct 2025What to knowThe CEO's statement; he added that labs are willing to pay even more, without a figure.SourceTechCrunchPress report
- Former investment bankers writing prompts and building financial models for OpenAI (Project Mercury)Price$150 per hour
per hour (paid to the expert)
WhenOct 2025What to knowRate from a person familiar with the effort; OpenAI said experts are paid through unnamed third-party suppliers, so the buyer's bill rate is higher.SourceFortune (citing Bloomberg)Press report
- Mercor's average pay for experts (October 2025)Price$85/hour
per hour (paid to the worker)
WhenOct 2025What to knowAgainst $105/hr in H1 2026 and a $109/hr homepage average in October 2026: average pay rose about 25% in a year.SourceContrary Research, Mercor memoPress report
- Google's planned 2025 spend with Scale AI for human-labeled training dataPrice$200 million this year
per year (2025, planned)
WhenJun 2025What to knowScale's largest customer at the time; Google planned to cut ties after Meta took 49% of Scale.SourceTechCrunch (citing Reuters)Press report
- A verified universal jailbreak of Anthropic's unreleased safety classifiers (bug bounty)Priceup to $25,000
per verified universal jailbreak
WhenMay 2025What to knowAgainst Constitutional Classifiers on Claude 3.7 Sonnet; the ceiling rose two-thirds in nine months. The programme concluded on 22 May 2025. Paid by the lab directly, no vendor.SourceAnthropic newsThe payer's own page
- Annotator writing IMO-level math problems for RLHF (Abaka AI cost model)Price20 USD per hour
per hour (paid to the worker)
WhenSep 2024What to knowA 2024 vendor cost model at one problem per hour; 2026 platform rates for formal-proof math run $90 to $200 an hour.SourceAbaka AI blog, LLM Data Cost BreakdownA vendor's published price guide
- Supervised fine-tuning (SFT) data, one constructed entryPriceabout 6 USD per entry
per entry (2,000 entries for 12,000 USD)
WhenSep 2024What to knowA vendor's 2024 worked example, older than the rest of this set.SourceAbaka AI blog, LLM Data Cost BreakdownA vendor's published price guide
- A verified universal jailbreak of Anthropic's unreleased safety classifiers (bug bounty)Priceup to $15,000
per verified universal jailbreak
WhenAug 2024What to knowInvite-only via HackerOne, on an unreleased safety system focused on CBRN and cyber. Paid by the lab directly, no vendor.SourceAnthropic newsThe payer's own page
- Mercor markup on contractor billing (February 2024)Priceapproximately 35%
markup on contractor pay
WhenFeb 2024What to knowThe earliest reported Mercor spread; the memo calls it both a markup and a take rate, and ties it to median BPO gross margins.SourceContrary Research, Mercor memoPress report
- Text labelling for ChatGPT safety, outsourced by OpenAI to Sama in Kenya (late 2021 contracts)Price$12.50 an hour
per hour, paid by OpenAI to Sama
WhenJan 2023What to knowOne of the few disclosed lab-to-vendor hourly prices; the workers took home $1.32 to $2 an hour, so the vendor kept most of the bill.SourceTIMEPress report
- Three OpenAI contracts with Sama for labelling workPriceabout $200,000 in total
per programme (three contracts)
WhenJan 2023What to knowCancelled early; OpenAI put the amount actually paid at about $150,000.SourceTIMEPress report
- Sama data labellers in Kenya working on OpenAI's ChatGPT safety projectPrice$1.32 to $2 per hour
per hour, take-home after tax
WhenJan 2023What to knowAgainst OpenAI's $12.50 an hour billed to Sama: the worker received 11 to 16% of the bill rate ($1.32 / $12.50 to $2 / $12.50).SourceTIMEPress report
Why labs pay for it
No archive contains an expert's judgment about a new problem. Most of what labs spend on data goes here: by our sum of the large vendors' newest reported annual revenue and run rates, roughly $8.1B a year.
RisingThe largest vendor of expert work passed a $2 billion gross run rate in mid-2026. Rates for generalists are falling while credentialed specialists hold $75 to $200 an hour.
- Professionals in any field
- Multilingual speakers
- Independent consultants
- Teams who can work to a brief
- A lawyer writing and grading legal tasks
- An accountant recording a month-end close on screen, with narration
- Conversations recorded in a chosen language
- Professional documents licensed by their author
- A credential and real experience in the field
- A scarce language
- Work in specialist software few people know
- Reliability over many accepted units
- General tasks anyone can do
- Work a model can already check by itself
Work made to a buyer's brief belongs to the buyer under the contract. Past work written for an employer or a client is theirs, not its author's, and cannot be sold by the person who wrote it.
Commissioned to a buyer's spec through AITraining.jobs, a job board for AI training work that has the same owner as this marketplace. People are paid per accepted unit.
AITraining.jobs, which shares an owner with this site, measures what AI training work pays per hour by kind of work from live job feeds.
Who buys it
SuperAnnotate
"Indicative $100K – $1M+", priced on rarity, depth and exclusivity.
Datacurve (Shipd)
Example bounties in the sample panels on its homepage run from $80 to $220 (read 16 Oct 2026; 2,847 registered contributors on its live counter). The panels show sample issues and challenges, not a rate card: "New quests launch every month across dozens of STEM categories".
Where expert work comes from
The whole supply map862,800
There were 862,800 physician and surgeon jobs in the US in 2025, a credentialed pool expert-data platforms recruit from.
Reaches buyers: Not reaching buyers (labour pool)
BLS Occupational Outlook Handbook, Physicians and Surgeons, 27 Aug 2026
~$60M to over 30K individuals every month
Handshake says its AI business reached about a $1B run rate and pays about $60M a month to over 30,000 people.
Reaches buyers: Contributor platform
Handshake job posting (Base10 job board), Fellow Experience Specialist (Contract), 15 Jun 2026
nearly five million experts
Mercor says it has nearly five million experts on its platform.
Reaches buyers: Contributor platform
900K+
Outlier says it has 900K+ AI trainers (master's holders, PhDs and graduates) and has paid experts more than $500M.
Reaches buyers: Contributor platform
more than 22%
In Q2 2026, Upwork's gross services volume from AI-related work grew more than 22% year over year, while total gross services volume fell 4% to $966.4 million.
Reaches buyers: Contributor platform (general freelance marketplace)
SEC EDGAR, Upwork Q2 2026 press release (8-K Ex. 99.1), 10 Aug 2026
about $500 million a year
The six largest Chinese AI labs spend about $500 million a year with American data-labelling companies, according to two founders in the sector who received market-size estimates from Tencent and ByteDance executives (Forbes).
Reaches buyers: US data vendors selling to Chinese labs
Forbes Colombia, Spanish edition of the Forbes article by Anna Tong (5 August 2026), 5 Aug 2026
What a buyer's counsel will ask
12 questions apply to expert work, and 7 of them block a sale until settled. Most data that fails to sell fails here, not on price.
- Does the company selling the data own it, and has someone with authority approved the sale?
Blocks a sale until settled
- Does any of the data belong to or describe your customers, and do your customer contracts and data processing agreements allow using it for anything other than serving them?
Blocks a sale until settled
- Does the data contain other companies' confidential information, such as partner documents, pricing, roadmaps or anything shared under an NDA?
Blocks a sale until settled
- Does any data come from health care, health insurance, or work you did for a health client under a business associate agreement?
Blocks a sale until settled
- Does the data include text, images, video or files created by others that you do not own (articles, reports, books, customer uploads, screenshots of other products)?
Blocks a sale until settled
- If the company is closing, insolvent or in bankruptcy, who now has legal authority to sell: the board, an assignee for the benefit of creditors, a bankruptcy trustee, or a Dutch curator?
Blocks a sale until settled
Need this made to a spec? Tell us what you need.
In this topic
What sells
- Start hereWhich data do AI labs want, and who holds it?Each kind of data the market prices, the businesses that hold it, what labs are asking for now and what they turn away.
- GuideWhat Are RL Environments? Who Builds, Buys and Prices ThemAn RL environment is a sandbox where an AI agent attempts tasks with tools and a verifier scores each try. Who builds them, who buys them and what they cost.
- GuideHow to Sell Videos to AI Companies: Prices, Buyers and RightsAI companies license video by the hour or minute. What buyers pay, which footage sells, and the rights you need in every frame before you sell a library.
- PageWhere data comes fromWho holds each kind of data, how much exists and how it reaches buyers.
- RegisterCompanies closingCompanies closing, sold or gone dark, and the records they leave.