Skip to content

Research report ยท September 2026 edition

The AI Training Rate Card

On 30 Sep 2026, for roles open to US residents, the best-paying platform advertised 2.09 times the rate of the worst-paying one in the typical kind of AI training work, and up to 4.33 times in Multilingual & Localization ($65 at Invisible against $15 at Welo Data).

Read the month's story: The AI Training Job Market in September 2026

Data window
1 Sep 2026 to 30 Sep 2026
30 days, UTC
Frozen
11 Oct 2026, 08:09 UTC
These figures will not change
Source
The public job boards of the AI training platforms this site tracks, read every day.
Method
Version 1. Definitions
CC BY 4.0, how to cite

Every figure on this page as CSV and JSON, sent to your work email.

Key findings

  1. 1

    Among roles open to US residents, the widest gap for the same kind of work is in Multilingual & Localization: Invisible advertises a median of $65 an hour and Welo Data $15 (4.33 times).

    Figure 1
  2. 2

    For roles open to US residents, the best-paying platform advertises 2.09 times the rate of the worst-paying one in the typical kind of work (9 kinds compared).

    Figure 1
  3. 3

    Across all roles, wherever they hire, the widest gap is in Multilingual & Localization: Invisible advertises a median of $65 an hour and Welo Data $10, 6.5 times as much, partly because the two hire in different countries.

    Figure 5
  4. 4

    Across all roles, wherever they hire, the best-paying platform advertises 2.82 times the rate of the worst-paying one in the typical kind of work (median across 9 kinds of work with two or more platforms compared).

    Figure 2
  5. 5

    For medicine & nursing roles, Mercor advertises a median of $100 an hour against $65 at OpenTrain AI (5 platforms compared).

    Figure 3
  6. 6

    For law roles, OpenTrain AI advertises a median of $270 an hour against $111 at Mercor (4 platforms compared).

    Figure 3
  7. 7

    For finance, accounting & economics roles, AfterQuery advertises a median of $108 an hour against $60 at Invisible (5 platforms compared).

    Figure 3
  8. 8

    For software & computer science roles, Mercor advertises a median of $80 an hour against $35 at OpenTrain AI (5 platforms compared).

    Figure 3
  9. 9

    For English, the most listed language, platform medians run from $21 (Mercor) to $65 (Invisible).

    Figure 4
  10. 10

    micro1 advertises the highest median rate in 5 kinds of work: Law, Medicine & Finance, Math & STEM, Writing & Content, Data Annotation & Labeling and Robotics & Physical-AI Data.

    Figure 6

Figure 1

Rate card for roles open to US residents

Like for like on geography: for roles a US resident can apply to, what does each platform advertise per hour?

Kind of workAll platformsOpenTrain AImicro1InvisibleMercorHighest over lowest
Math & STEM$85$95$115$65$95$100$45n/an/an/an/a2.56x
Law, Medicine & Finance$80$90$115$65$90$83$35$93n/an/an/a3.29x
Coding & Software Evaluation$70$75$70$65$80$75n/an/an/an/an/a1.23x
Agentic Tasks & RL Environments$60$60n/an/an/an/an/an/an/an/an/an/a
AI Evaluation & Rating (RLHF)$60$32$65$45$70n/an/an/an/an/an/a2.19x
Writing & Content$60$60$60n/an/an/an/an/an/an/an/a1.00x
Multilingual & Localization$40$20$38$65$41n/an/a$18$15n/a$404.33x
Data Annotation & Labeling$34$23$48n/an/an/an/an/a$34n/an/a2.09x
Transcription, Audio & Speech$30$28$20$36n/a$30$23$27$27$35n/a1.80x
Red-Teaming & AI Safety$28n/an/an/a$21n/an/an/an/an/an/an/a
Robotics & Physical-AI Data$15$23$15n/an/an/an/an/an/an/an/a1.53x

Sample. 1,482 roles open to US residents, open at the end of 30 Sep 2026, that state an hourly rate. A platform's median is shown where at least 5 of its roles in the row state one; n/a otherwise.

Advertised pay: the hourly figure a posting states (the middle of a stated range). What a worker is paid can differ with tiers, bonuses and rejected tasks.

Columns are the 10 platforms with the most published medians in this table; "All platforms" counts every platform's priced roles in the row.

Only roles a resident of the United States may apply to (open worldwide, listing the US, or not excluding it), so a platform hiring mostly in lower-cost countries is compared on the same ground.

Source. AITraining.jobs Research, The AI Training Rate Card, September 2026 edition.

Figure 2

Rate card by kind of work

For each kind of AI training work, what does each platform advertise per hour?

Kind of workAll platformsOpenTrain AImicro1InvisibleMercorHighest over lowest
Law, Medicine & Finance$78$90$115$65$90$83$93n/a$35n/an/a3.29x
Coding & Software Evaluation$65$27$70$65$79$75n/an/an/an/an/a2.93x
Math & STEM$65$95$115$65$95$100n/an/a$45n/an/a2.56x
Agentic Tasks & RL Environments$60$60n/an/an/an/an/an/an/an/an/an/a
AI Evaluation & Rating (RLHF)$60$30$65$53$70n/a$19$20n/an/an/a3.68x
Writing & Content$60$60$60n/an/an/an/an/an/an/an/a1.00x
Red-Teaming & AI Safety$39n/an/an/a$33n/an/an/an/an/an/an/a
Multilingual & Localization$37$20$38$65$38n/a$18$10n/an/a$406.50x
Transcription, Audio & Speech$30$23$20$36n/a$30$25$27$23$35n/a1.80x
Data Annotation & Labeling$28$17$48n/an/an/an/a$32n/an/an/a2.82x
Robotics & Physical-AI Data$15$15$15n/an/an/an/an/an/an/an/a1.00x

Sample. 1,967 roles, open at the end of 30 Sep 2026, that state an hourly rate. A platform's median is shown where at least 5 of its roles in the row state one; n/a otherwise.

Advertised pay: the hourly figure a posting states (the middle of a stated range). What a worker is paid can differ with tiers, bonuses and rejected tasks.

Columns are the 10 platforms with the most published medians in this table; "All platforms" counts every platform's priced roles in the row.

Source. AITraining.jobs Research, The AI Training Rate Card, September 2026 edition.

Figure 3

Rate card by expert field

For each profession or discipline named in a title, what does each platform advertise per hour?

Expert fieldAll platformsmicro1OpenTrain AIMercorAfterQueryHighest over lowest
Psychology & mental health$150$150$150n/a$88n/an/a1.70x
Law$125$130$270$111$125n/an/a2.43x
Physics & astronomy$95$150$120$95n/an/an/a1.58x
Finance, accounting & economics$79$73$73$80$108$60n/a1.80x
Business, sales & consulting$75$65$68$90$85n/an/a1.38x
Engineering (non-software)$75$70$70$78$98$65n/a1.51x
Medicine & nursing$69$65$65$100$80$65n/a1.54x
Biology & life sciences$65n/an/an/an/a$65n/an/a
Chemistry$65n/an/an/an/a$65n/an/a
Mathematics & statistics$65n/a$38n/an/a$65n/a1.71x
Software & computer science$65$75$35$80$78$65n/a2.29x
Education & teaching$61$115n/an/a$54n/an/a2.13x
Humanities & social sciences$60n/an/a$55n/a$50n/a1.10x
Design & visual arts$55$44$50$75n/an/an/a1.70x
Writing, editing & journalism$45$60$40n/a$69n/a$233.00x
Voice acting, music & sound$36n/an/a$45n/a$36n/a1.25x

Sample. Roles whose title names an expert field, open at the end of 30 Sep 2026, that state an hourly rate. A platform's median is shown where at least 5 of its roles in the row state one; n/a otherwise.

Advertised pay: the hourly figure a posting states (the middle of a stated range). What a worker is paid can differ with tiers, bonuses and rejected tasks.

Columns are the 6 platforms with the most published medians in this table; "All platforms" counts every platform's priced roles in the row.

Source. AITraining.jobs Research, The AI Training Rate Card, September 2026 edition.

Figure 4

Rate card by language

For each language named in a title, what does each platform advertise per hour?

LanguageAll platformsInvisibleOpenTrain AIWelo DataMercorHighest over lowest
Indonesian$63$65$63n/an/an/an/an/a1.03x
Russian$45$65n/an/an/an/an/an/an/a
Swedish$38$65n/a$27n/an/an/an/a2.41x
Arabic$37$65n/a$27n/an/an/an/a2.41x
Dutch$37$65n/a$15n/an/an/an/a4.33x
German$36$65$23$33$50n/an/an/a2.83x
Japanese$36$65$29n/an/an/an/a$482.24x
French$35$65$23$32$50n/an/an/a2.83x
Spanish$34$65$23$33$40n/a$20n/a3.25x
Hindi$33$65$21n/an/an/an/an/a3.10x
Korean$31$65$25n/an/an/an/an/a2.60x
English$30$65$32$27$21$30n/an/a3.10x
Portuguese$30$65$23$27n/an/an/an/a2.83x
Italian$28$65$23$17n/an/an/an/a3.82x
Chinese (Mandarin)$26$37$20$15n/an/an/an/a2.47x

Sample. Roles whose title names a language, open at the end of 30 Sep 2026, that state an hourly rate. A platform's median is shown where at least 5 of its roles in the row state one; n/a otherwise.

Advertised pay: the hourly figure a posting states (the middle of a stated range). What a worker is paid can differ with tiers, bonuses and rejected tasks.

Columns are the 7 platforms with the most published medians in this table; "All platforms" counts every platform's priced roles in the row.

Source. AITraining.jobs Research, The AI Training Rate Card, September 2026 edition.

Figure 5

How much the platform changes the rate for the same work

For the same work, how far apart are the best- and worst-paying platforms?

WorkHighestLowestHighest over lowest
Multilingual & LocalizationInvisible $65Welo Data $10
6.50x
DutchInvisible $65Welo Data $15
4.33x
ItalianInvisible $65Welo Data $17
3.82x
AI Evaluation & Rating (RLHF)Mercor $70Braintrust $19
3.68x
Law, Medicine & Financemicro1 $115Alignerr $35
3.29x
SpanishInvisible $65Alignerr $20
3.25x
EnglishInvisible $65Mercor $21
3.10x
HindiInvisible $65OpenTrain AI $21
3.10x
Writing, editing & journalismAfterQuery $69Alignerr $23
3.00x
Coding & Software EvaluationMercor $79OpenTrain AI $27
2.93x
FrenchInvisible $65OpenTrain AI $23
2.83x
GermanInvisible $65OpenTrain AI $23
2.83x
PortugueseInvisible $65OpenTrain AI $23
2.83x
Data Annotation & Labelingmicro1 $48OpenTrain AI $17
2.82x
KoreanInvisible $65OpenTrain AI $25
2.60x
Math & STEMmicro1 $115Alignerr $45
2.56x
Chinese (Mandarin)Invisible $37Welo Data $15
2.47x
LawOpenTrain AI $270Mercor $111
2.43x
ArabicInvisible $65Welo Data $27
2.41x
SwedishInvisible $65Welo Data $27
2.41x
Software & computer scienceMercor $80OpenTrain AI $35
2.29x
JapaneseInvisible $65OpenTrain AI $29
2.24x
Education & teachingmicro1 $115AfterQuery $54
2.13x
Transcription, Audio & SpeechInvisible $36micro1 $20
1.80x
Finance, accounting & economicsAfterQuery $108Invisible $60
1.80x
Mathematics & statisticsInvisible $65OpenTrain AI $38
1.71x
Design & visual artsMercor $75micro1 $44
1.70x
Psychology & mental healthmicro1 $150AfterQuery $88
1.70x
Physics & astronomymicro1 $150Mercor $95
1.58x
Medicine & nursingMercor $100OpenTrain AI $65
1.54x
Engineering (non-software)AfterQuery $98Invisible $65
1.51x
Business, sales & consultingMercor $90micro1 $65
1.38x
Voice acting, music & soundMercor $45Invisible $36
1.25x
Humanities & social sciencesMercor $55Invisible $50
1.10x
IndonesianInvisible $65OpenTrain AI $63
1.03x
Robotics & Physical-AI Datamicro1 $15OpenTrain AI $15
1.00x
Writing & Contentmicro1 $60OpenTrain AI $60
1.00x

Sample. Groups in which at least two platforms each have 5 or more open roles stating an hourly rate, on 30 Sep 2026.

A spread compares platform medians, not individual roles: one platform can pay a wide range within the same work.

Across all roles a spread also reflects where each platform hires: many set lower rates in lower-cost countries. The rate card for roles open to US residents compares like for like.

Source. AITraining.jobs Research, The AI Training Rate Card, September 2026 edition.

Figure 6

Which platform advertises the most for each kind of work

Which platforms pay the top rate in the most kinds of AI training work?

PlatformKinds of work where it pays the mostWhich
micro1
5
Law, Medicine & Finance, Math & STEM, Writing & Content, Data Annotation & Labeling and Robotics & Physical-AI Data
Invisible
2
Transcription, Audio & Speech and Multilingual & Localization
Mercor
2
Coding & Software Evaluation and AI Evaluation & Rating (RLHF)

Sample. Kinds of work with at least two platforms each advertising an hourly rate on 5 or more open roles, on 30 Sep 2026.

Source. AITraining.jobs Research, The AI Training Rate Card, September 2026 edition.

For labs, data vendors and researchers

Buy the data behind this report

The figures above are a summary. Behind them are records we collect ourselves, every day:

  • 10,289 roles tracked, each with its history

    Every role on 29 job boards, read daily since 13 Jul 2026: its pay and place, and the first and last day it was open.

  • 6,200 apply clicks in 28 days

    Which roles people actually apply to, by field, pay and kind of work. It happens on this site, so nobody else can count it.

  • 39 lab and vendor hiring boards

    The job boards of 25 AI labs and 14 human-data vendors, read every day: who is hiring for human data, and who is hiring to buy it.

What you can buy

  • The records

    The rows behind any figure, with their history, not only the totals.

  • A regular feed

    The same figures every week or month, as files.

  • A custom cut or benchmark

    Your countries, languages or fields, or your own platform against the market.

Method and definitions

The public job boards of the AI training platforms this site tracks, read every day. Every figure is computed by one database query over the window shown, the same query for the live report and for every edition. Each figure states its sample beneath it.

Platform median
The median advertised hourly rate of a platform's open roles in a group that state one, published only with at least 5 such roles.
Like for like
The rate card for roles open to US residents compares platforms on the same geography: many set lower rates in lower-cost countries.
Highest over lowest
The best-paying platform's median divided by the worst-paying platform's, among platforms with a published median.
Advertised pay
The hourly figure a posting states (the middle of a stated range). Medians need at least 8 roles stating a rate; the middle half runs from the 25th to the 75th percentile.
Kind of work
Each role is classified from its title and description into one of the twelve kinds of AI training work this site tracks, from AI evaluation and coding to multilingual and audio work. Every kind links to its own page.
Location rule
Read from each posting's location text and, where that names no country, from what the posting's own description states: open worldwide, worldwide except listed countries, only listed countries, or no rule stated.
Window
Live: the last 28 complete days, in UTC. Edition: one calendar month, in UTC.

What this data cannot tell you

  • Platforms that recruit without a public job board are not counted. Several large ones invite workers to projects directly.
  • A listing is not a hire. Some platforms keep evergreen roles open for months, and some post one listing per language or country.
  • Pay is what postings advertise, not what workers are paid after tiers, bonuses or rejected tasks.
  • A platform median covers the roles it lists in that group; two platforms' roles in one group can still differ in seniority and difficulty.

How the job feeds are read and roles are classified: site methodology. How fresh, complete and consistent the data is, checked every day: data quality and coverage.

Editions

The live report covers the last 28 complete days and changes every day. Each complete month is also published as an edition, frozen two days after the month ends, so a figure you cite from an edition never moves.

Cite and reuse

Free to quote, chart and republish under CC BY 4.0 with a link to this page. This edition is frozen, so its figures will read the same later.

Citation

AITraining.jobs (2026). The AI Training Rate Card, September 2026 edition. https://www.aitraining.run/research/ai-training-rate-card/2026-09

BibTeX

@misc{aitrainingjobs_ai_training_rate_card_2026_09,
  author = {{AITraining.jobs}},
  title = {The AI Training Rate Card, September 2026 edition},
  year = {2026},
  howpublished = {\url{https://www.aitraining.run/research/ai-training-rate-card/2026-09}},
}
Data: every figure as CSV (one line per value) and JSON, sent to your work email.Pay and open-roles data

More research

All 78 research reports