Model PMs earn $345K median base, 60% more than PMs outside AI

aakashgupta · x · 2026-10-17

Analyzing 1,177 open PM postings from company career pages (Oct 8–13), the author sorted salaries by actual PM work type (base only, no equity):

Model PM: $345K median (46 open roles)
AI-native PM: $258K (79 postings)
AI platform PM: $231K (140 postings)
Applied AI PM: $222K (297 postings)
PM outside AI: $216K (615 postings)

Model PMs build the models themselves alongside research teams. Six recurring jobs: decide what the model gets better at, turn "better" into an eval, read failed-run transcripts, write behavioral specs, make the ship call, launch to developers.

Same-level comparisons show the premium: mid-level +86% ($371K vs $200K), senior +73%, staff/principal +35%. Top posted ranges: OpenAI Core Models PM up to $490K, ByteDance AI Data and Safety Senior PM $462K, Anthropic Safeguards PM $460K, NVIDIA Nemotron $426K. OpenAI's median PM total comp was $860K including equity.

Notably, only 1 of 46 postings requires a PhD, 8 ask for Python/SQL, and 76% mention evals — the most common skill. Madhu Guru (ex-Google Gemini/Veo product lead) estimates fewer than 75 PMs worldwide have AI-native product depth. The author suggests building eval skills tonight: pick a task the model fails at, write 10 failing cases, label correct answers, compare two models.

Original post →

Model PMs earn $345K median base, 60% more than PMs outside AI

More from Companies & People