US midterms · Tue 3 Nov 2026 · loading
On election night we run every US midterm race through Jevlin: one call, 6 of its search engines, every few minutes from the time polls close. We watch which engine inside it finds each result first. A race scores when an engine's results name the winner of AP's call, after a person has checked that call. This is the provider ranking: who knew first.
Searching stopped 10 Nov. AP called five races after that (Mississippi Senate, Kansas governor, Maine 2nd, Colorado 4th, Alaska governor); in those, an engine that had not named the winner by 10 Nov counts as not found.
Median minutes from poll close to finding AP's winner, over called races. Wrong = races where the engine's results named a different candidate.
Answers come only from the measurements on this page, and the dashboard rearranges to show them. Ask doesn't discuss candidates, parties or predictions.
No races match these filters.
These are the 156 races we time: every Senate and Governor race and 85 competitive House races. For every other race, see AP's full results.
Over called races only. Minutes from poll close to the first five-minute round whose results named AP's winner. Wrong counts races where an engine's results named a different candidate; it sits beside every row. Tap a column to sort.
"Jevlin, all engines" is the earliest of the six engines from the same calls, so it costs nothing extra: its cost is the whole call. Cost is what each engine costs per call through Jevlin, as billed tonight (no bulk pricing with any provider).
–Not an official results source. Race calls are from AP, checked by our team before they appear here.