OpenAI rolls hard spend caps out to all API accounts this week
A monthly ceiling that hard-stops requests with a 429 once you hit it — not just a soft alert. Plus: Cursor's Auto mode now picks a fresh model per request.
Copy markdown
Set a ceiling, get a 429 — not an overage email
Hard spend limits are rolling out to every OpenAI API account this week: set a monthly dollar cap and requests return HTTP 429 the moment tracked spend crosses it, replacing the old soft alerts that let bills sail past your budget. It's the first real circuit-breaker for agent and eval workloads, where a single loop can fan out into thousands of calls before anyone notices.
Cursor's Auto mode now routes per request
Cursor's new Router picks a model for each individual request instead of pinning Auto to one, across three modes: Intelligence (max quality), Balance (a daily driver), and Cost (fewer tokens). Team admins can lock the mode or trim the model allowlist — so your Auto-mode quality and spend may now hinge on a setting someone else controls.