Post-Training Frontiers.

Model trainingPost-training

OpenAI's public role description says Post-Training Frontiers trains frontier agents with large-scale reinforcement learning and builds the systems used to run and evaluate that work.

Public source

Public sources are provided for context; no recent verification date recorded.

Compare teams

Community intelligence

Core decision signals

1 community contributor. Work-life balance, reorg frequency and technical depth are shown first. Each dimension has its own sample; skipped answers do not count.

Work-life balance

Not enough data

Reorgs, past 12 months

Not enough data

Technical depth

Not enough data

More community dimensionsModel ownership, on-call, publishing, would join again and compensation

Model ownership

Not enough data

On-call intensity

Not enough data

Publishing freedom

Not enough data

Would join again

Not enough data

Compensation privacy

Annual compensation

Self-reported and separate from public job-posting ranges. Base and total each need three eligible reports.

Base salary

0 reports collected

3 more needed to show the range

Total annual compensation

0 reports collected

3 more needed to show the range

Read the sample, not a score.

Results appear only after at least three qualifying contributors answer that specific question. Compensation and notes remain separate from public facts.

Ask or browse team questions