Google's ERA-built model ranks #1 of 39 in the CDC FluSight 2025–26 flu-hospitalization forecasting challenge
Google reported on Sept 30, 2026 that its best forecasting model ranked first of 39 eligible models in the CDC's FluSight challenge for the 2025–26 US flu season (October–May). FluSight forecasts weekly hospital admissions now and up to three weeks ahead. The forecasts were built with Empirical Research Assistance (ERA), Google's AI tool that writes and optimizes scientific code.
Key facts
- Challenge: CDC FluSight, 2025–26 season; end-of-season analysis published Sept 30, 2026
- Google: 'of 39 eligible models, Google's best matched the season's observed hospital admissions' (current-week and up to three-weeks-ahead forecasts)
- Built with ERA (Empirical Research Assistance), the AI system behind the Gemini for Science computational-discovery tools
Science result
- Field
- medicine / epidemiological forecasting
- Problem
- Forecasting weekly US influenza hospital admissions (CDC FluSight)
- Result
- Ranked first of 39 eligible models in the 2025–26 season.
- AI system
- ERA (Empirical Research Assistance)
- Human role
- AI-assisted: ERA generated and optimized forecasting code; Google researchers submitted the forecasts
- Verification
- CDC FluSight end-of-season evaluation (as reported by Google)
- Status
- confirmed
What happened
Google entered FluSight with forecasting models developed through ERA. At season's end its best model ranked first among the eligible entries, which come from academic and public-health teams.
Why it matters
It is a real-world, prospectively scored test of AI-generated scientific software, not a retrospective benchmark.
Changelog
- 2026-10-01: created (Google blog check)
Related events
- Google launches 'Gemini for Science' at I/O 2026: Co-Scientist, AlphaEvolve and ERA become products ★★★
Sources (1)
id: 2026-09-30-google-era-flusight-flu-forecasting · updated 2026-10-01 · open in the interactive timeline