Skip to content
← Models
Open weightsMicrosoft Research/10 trustMetrics-checked

Microsoft Aurora

Open foundation model for the Earth system — weather, air quality, waves, and cyclones.

Visit official page →

About

Aurora is Microsoft Research's foundation model for atmospheric prediction, first described in a May 2024 preprint and published in Nature in May 2025. Rather than training one model per forecasting task, the team pretrained a single encoder-based model on over one million hours of atmospheric data — satellites, radar, weather stations, and simulations, described as the largest such collection ever assembled — then fine-tuned it for medium-range weather, air quality, ocean waves, and tropical cyclones.

Fine-tuned Aurora beat existing numerical and AI systems on about 91% of medium-range forecasting targets at 0.25-degree resolution, and a specialized version already powers hourly forecasts in MSN Weather. Notably, Microsoft released both code and weights openly (MIT-licensed checkpoints on Hugging Face, e.g. aurora-0.25-pretrained), making it one of the rare frontier-class scientific models the community can actually run and build on.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

May 1, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

weather-forecastingair-qualityclimate-researchdisaster-preparedness

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 80; Trending score 74/100 from researcher assessment.

Sources