Neither company publishes a historical monthly series, so this report assembles representative point-in-time snapshots of company-wide advertised open roles from independent job-board trackers and articles, January through September 2026 (latest observations through September 22). Coverage is 17 of 18 company-month cells: February has no resolvable Anthropic count, and the February OpenAI figure of 800 comes from a single low-confidence source.
Every independent observation in the research, plotted by month. Filled dots are the selected representative values; open dots are alternative crawls. The spread shows these are snapshots, not monthly averages — crawl timing and duplicate multi-location postings drive most disagreement. One March outlier (an OpenAI announcement of 3,500 planned roles, not live listings) is capped at the top of the scale. Hover or tap a dot for its source.
| Month | OpenAI | Source | Anthropic | Source |
|---|
Company-wide advertised open roles at OpenAI and Anthropic, point-in-time snapshots per month, January–September 2026 (latest through Sep 22). 103 distinct observations from 114 evidence rows across independent job boards, trackers, and articles; one representative value selected per company-month (17 of 18 cells; February Anthropic unresolved, February OpenAI low-confidence). Counts are live advertised openings, not hires, headcount, or monthly averages; differences between trackers reflect crawl timing and deduplication of multi-location postings. Research scanned over 1,200 unique web results across two broad search passes plus targeted searches. Alternative observations appear only in the spread chart for space.