Downloads
- Pay by field (CSV, 14 rows)
- Pay by platform (CSV, 15 rows)
- Pay by country you can work from (CSV, 15 rows)
- Pay for roles that need another language (CSV, 15 rows)
Each file has one row per group: the number of open roles, the number of distinct postings that state an hourly rate, and the median, 25th, 75th and 90th percentile and highest hourly rate in US dollars. The files are rebuilt every hour from the roles listed at the time; these figures are from 3 November 2026.
मासिक रिपोर्ट
Each month's figures, frozen once the month ends:
How the figures are made
Only roles that state an hourly rate count. Each distinct posting counts once, "up to" ceilings and paid research studies are left out, and a range counts at its midpoint. The methodology has the details.
Reusing the data
These aggregate tables are licensed under Creative Commons Attribution 4.0: you can copy, share and adapt them, including commercially, as long as you credit aitrainers.jobs and link to this page. The licence covers these tables only, not the job listings on the site.
To search the open roles themselves from your own code or an AI assistant, see the search API and MCP server.
Suggested credit: aitrainers.jobs, AI training pay data, 3 November 2026, https://aitrainers.jobs/open-data
Archived versions with a DOI
Monthly snapshots of these tables, and of the trend series on the trends pages, are archived on Zenodo with a permanent DOI, so a figure you cite can always be found as it was: doi.org/10.5281/zenodo.23596529 (always the newest version; each version has its own DOI too). The same files are in a public GitHub repository and on Kaggle, ready to load in a notebook.
To cite an archived version: aitrainers.jobs (2026). AI training jobs: pay and hiring data [Data set]. Zenodo. https://doi.org/10.5281/zenodo.23596529