Live pay ranking · October 2026
The highest-paying AI training platforms, ranked by live rates
Compare advertised pay for specialist, annotation and evaluation work, then check each role's qualifications and country restrictions. A high platform ranking does not mean every role suits you.
16 platforms publish enough distinct rates to rank by median top advertised hourly rate. One pay band counts once (methodology). The site tracks 3,234 open roles overall, including work outside this ranking. See the full pay index for the wider picture. 24 other platforms are listed separately because their feed evidence does not meet this ranking's sample requirement.
Ranked by typical top rate (live feeds)
One of the largest AI-work marketplaces; $10B valuation (Oct 2025). Rates and roles verified from their public listings.
Checked 2026-09-19: San Francisco company that supplies coding, STEM and expert training data to frontier AI labs, according to its own help centre. Its public board at work.turing.com/jobs showed 226 roles, nearly all remote contract work. Its on-site and hybrid office roles are not carried here. Applying needs a Turing account
YC-backed expert data platform. Its public board showed 126 open listings on 2026-08-26, each printing its own rate, no account needed.
Fast-growing Scale AI challenger ($500M valuation). Apply once, AI interview, then matched to projects.
PhD and graduate-focused AI training fellowship. Its own board showed 203 open roles on 2026-08-26, each on its own public page with a rate, no account needed.
Alignerr advertises each job many times over. Its board showed 5,530 listings on 2026-08-26 but only 469 separate jobs. Some are copied once per city, with the city name dropped into a single sentence of an otherwise identical posting, and every listing still says Remote. Others are simply repeated, word for word or lightly reworded. This directory lists each job once. Listings carry no date, and some workers report slow payment resolution, so apply with that in mind.
- 7.xAI Specialist AI Tutorsmedian $75/hr · $75–$100 typicaltop $100/hr · 5 distinct rates (25 listings)
Direct contractor roles at an AI lab. Generalist hiring paused in 2025; specialist tutors (STEM, medicine, finance) still recruited.
AI-native expert panel blending market research with frontier-AI evaluation work. The public board at terac.com/opportunities is readable without an account. Its board showed 73 open opportunities on 2026-08-26. Only 17 of them pay by the hour, 46 pay a one-time fee and 10 pay per task.
Checked 2026-10-04: SME Careers describes itself in its postings as an AI data services company and subsidiary of SuperAnnotate, and on 2026-10-03 its site footer named SME Solutions Inc. Its public board at sme.careers/jobs showed 261 roles to a signed-out visitor, each with its rate, behind 1,695 country postings, and 1,598 of those postings were remote contract work. 955 of the 1,695 said in their own text that there was no immediate project for the role, and this directory does not list them. On 2026-10-03 a role's own panel listed one application step, an AI interview for that role, and applying sent a signed-out visitor to sign in or sign up first. Some postings say selection then adds a domain task and a recruiter interview
- 10.Meridial (Invisible)median $45/hr · $28–$65 typicaltop $200/hr · 29 distinct rates (762 listings)
Invisible Technologies' expert-contractor network. 800+ open freelance AI-trainer projects.
Checked 2026-09-12: public roles cover annotation, coding, languages and domain expertise. The board remains the source of availability, including older listings that are still advertised
Checked 2026-10-05: clinical and research roles for doctors, scientists and regulatory specialists. Closed roles stay online marked closed and leave the board
Checked 2026-09-12: public roles span annotation, model evaluation and video capture. The open board determines availability because a still-listed role carried an expired validThrough date.
Checked 2026-09-12: Trustpilot reviewers describe long screening processes and unsolicited contact after applying. These are individual reports about the marketplace, not verified outcomes for every employer
Checked 2026-10-07: Lionbridge's AI Data Training team posts freelance transcription, annotation and data collection projects on its own Lever board. Its site names Lionbridge Technologies, LLC
Checked 2026-10-04: RWS's public Lever board carried 47 postings, all on its TrainAI team. Of those, 20 were posted in September 2026, the newest on 2026-09-29, among them 8 Legal Annotators postings, one per language or script, posted on 2026-09-16 with a stated duration to the end of October. Another 15 date from before June 2026, including the 8 Search Engine Evaluator roles posted on 2024-10-14 and a talent-pool posting from 2025-03-14, which this directory does not list. Lever files 46 as temporary or contract roles and 1 as a part-time employee role in the United States
Compare two platforms head to head
Other platforms: advertised or reported pay
These platforms do not currently meet the ranking's sample requirement. Some have a public feed with too few distinct rates, while others have no readable feed. The descriptions below are not ranked medians.
- Sovrano AIChecked 2026-09-20: every published role carries a structured rate. 21 were hourly, at 60 US dollars an hour for the slide and deck expert roles, 12 euros an hour for the native-speaker evaluator roles, 8 to 12 euros an hour for the speech data roles and 5 to 15 US dollars an hour for robotics annotation. One paid 500 US dollars per approved task. Four paid 1,000 euros a month, stored as an annual salary the way this directory stores every monthly figure. Eleven of the hourly roles state in their own payment terms that the rate is for a pilot and that production rates are still to be confirmed, so the euro figures on those may change. The roles at 8 euros an hour, and the monthly ones, fall under this directory's 10 dollar floor, so they are not listed
- TransPerfect DataForcePublished payment is often per accepted recording, photo, story or study session. An hour of accepted media is a task unit, not an hourly wage.
- Welo Data (Welocalize)Rates in listings, strongly geo-dependent (≈$2–$13/hr by country)
- Pasiflora AIRates come from each open role's headline pay field. Approved-participant and recorded-video payments stay task units. Separate proposal fees remain in the role description.
- G2iExplicit hourly ranges appear in the role descriptions. Weekly workload hours are kept separate from compensation.
- Fleet AIFellow roles publish typical hourly ranges. Final compensation depends on domain, experience, location and engagement.
- InnodataContributor roles publish hourly pay. Repeated state copies with the same title and compensation collapse to one role. Canadian rates remain in CAD.
- iMerit ScholarsThe public feed publishes explicit hourly rates alongside low-paying country variants. Roles below the directory floor and amounts without a currency or unit are omitted.
- MindriftAdvertised $15–$100+/hr depending on domain; payouts twice monthly (Payoneer/PayPal)
- CrowdGen by AppenStructured per-task payments published per project (e.g. $15–$50 per recording set)
- OneForma (Centific)Publishes no amount anywhere. States only that you see your rate before accepting a project, and that payment carries no minimum threshold
- Outlier (Scale AI)Public opportunities advertise up-to hourly ceilings. Rates depend on assessment, location and project; onboarding or non-core work can pay less
- DataAnnotationPublic recruitment pages advertise up to $25–$50/hr for generalists, up to $75–$150+/hr for software engineers, and role-specific specialist figures. These are advertised ceilings; assessment results and available projects determine actual work and pay
- EthosPublishes an exact rate and a weekly cap on each opportunity page, seen from $50/hr up to $225/hr depending on the role
- Babel AudioPublishes a rate on every role, but the unit is an hour of audio delivered, not an hour worked. Conversation roles pay a flat rate per recorded hour of talking, which is close to an hourly wage. Transcription pays up to a stated figure per audio hour transcribed and voice acting a flat figure per recorded hour of finished audio, and both take several working hours to produce, so they are shown per task here and never as an hourly rate
- HumanSignalChecked 2026-10-04: its contributor page shows example rates for US residents, $22 to $45 an hour for AI response evaluation and $15 to $35 for data collection, both paid per accepted task, and $60 to $180 for domain expert review. It says each project shows its rate before work starts. Its Greenhouse postings give other figures in their text, from up to $6 an hour on several country data collection postings to up to $100 an hour for US candidates on its remote one.
- ProxionChecked 2026-10-04: its experts page and its one posting advertise $75 to $300 an hour, set by specialization, seniority and project complexity, and say pay can rise for reviewers and project leads. It says the technical test is paid and that contractors invoice and are paid bi-weekly. It publishes no individual openings with their own rate, so no rate here is read from a listing.
- NobleStark HatchChecked 2026-10-04: its Hatch page lists five tracks with an hourly band each, $120 to $200 for agentic software engineering, $80 to $160 for code generation and review, $150 to $300 for cybersecurity and red-team work, $100 to $250 for math and formal reasoning, and $140 to $220 for custom eval and benchmark authoring. It says the trial task is paid but not how much, and that payouts are weekly. It publishes no individual openings, so no rate here is read from a listing.
- LuelChecked 2026-10-04: its contributor page showed six tasks with a rate each, $12 an hour for American English conversations among three people and $7.50 an hour for partner conversations in Uyghur, Tibetan, Hmong, Jamaican Creole and Guarani. Its terms say conversation pay is counted per participant by duration, that no rate is locked when you record or upload, and that the rate paid is the one in force when a submission is approved. Its contributor page says payouts are processed daily.
- PeritChecked 2026-10-04: its recording page pays per approved audio hour, $20 for read sessions, $22 for prompted speech, $24 for everyday rooms, $28 for conversations and $20 to $32 for accents it is short on. An audio hour is recorded audio, not time worked, and its own example puts a 20 minute session at about $7. Its careers page shows $17 to $45 per audio hour on some transcription and alignment roles, and $210 to $365 a month on its India rater and annotation contracts, which its job data states as 20,000 to 35,000 Indian rupees.
- WorkadaChecked 2026-10-07: every public role page publishes a USD hourly rate, from 17 dollars for data labeling to 100 dollars for finance and life science roles. Pay is sent three times a week through Stripe
- VettoChecked 2026-10-07: most roles pay a fixed US dollar fee per approved task or deliverable, from 30 to 60 dollars, paid in local currency. The hourly figures beside the fees are Vetto's own estimates and are not shown as rates here
- 10x TeamChecked 2026-10-07: every AI trainer posting shows a euro hourly range, from 25 euros at the low end to 375 at the high end, for 8 to 20 hours a week of freelance work
- MeredenChecked 2026-10-10: its site lists 33 roles, and 32 of them show an hourly band in US dollars, from $15 to $30 for robotics reviewers and driving experts up to $100 to $150 for olympiad medallists. Medical experts and STEM PhDs are shown at $80 to $140, legal experts at $70 to $130, and its general mathematics, coding and STEM evaluator roles at $20 to $40. The 33rd invites you to propose a role of your own and names no rate. Its role pages say the vetting task is paid and that payouts go out weekly. It publishes no dates on its roles, so no rate here is read from a dated listing.
Rankings recompute automatically as feeds refresh. Some outbound links are referral links, and they are never a ranking factor (disclosure). Browse all live roles sorted by pay →