New AI model advances the prediction of weather uncertainties and risks, delivering faster, more accurate forecasts up to 15 days ahead. Weather affects everyone, guiding daily choices, safety measures, and overall lifestyle. As climate change intensifies the frequency and severity of extreme events, having forecasts that are both accurate and trustworthy becomes ever more crucial. Yet, no forecast can be perfect, and uncertainties grow as forecasts reach further into the future. Because of this inherent limit, scientists and weather agencies rely on probabilistic ensemble forecasts, which present a range of likely weather scenarios rather than a single outcome. Ensemble forecasts are inherently more informative for decision-makers, offering a fuller picture of potential conditions in the coming days and weeks and indicating how likely each scenario is.
Today, in a peer-reviewed report, researchers introduce GenCast, a new high-resolution AI ensemble model with a resolution of 0.25 degrees. GenCast promises improved forecasts for both routine day-to-day weather and extreme events, outperforming the leading operational system of record up to 15 days ahead. The authors announce their intention to release the model’s code, weights, and forecasts to support the broader weather forecasting community and encourage widespread adoption and validation.
Understanding Probabilistic Ensemble Forecasts and Weather Uncertainty
Weather is an inherently complex system governed by nonlinear dynamics, countless interacting processes, and a changing climate. Even with advanced physics-based models and dense observational networks, predicting the exact state of the atmosphere at a specific future time remains elusive beyond short time horizons. This fundamental limitation underpins the shift from single deterministic forecasts to probabilistic ensemble approaches.
An ensemble forecast consists of multiple model runs, each initiated with slightly different starting conditions, physics configurations, or parameter choices. The dispersion among these runs—the ensemble spread—encodes forecast uncertainty. When the ensemble converges toward a common outcome, confidence in that outcome increases; when it splinters into a wide array of possibilities, uncertainty grows, and risk assessments must reflect that breadth.
The value of ensemble forecasts lies in several dimensions. First, they quantify uncertainty directly, enabling probabilistic interpretations such as the likelihood of precipitation, the probability of exceeding a temperature threshold, or the chance of a heavy rainfall event. Second, they aid risk management by highlighting worst-case or best-case scenarios, supporting decisions in sectors ranging from agriculture and energy to transportation and emergency management. Third, ensembles improve communication with end users by translating complex physics into actionable information—forecast probabilities, confidence bands, and scenario narratives that reflect real-world risks.
High-resolution ensembles are particularly valuable because small-scale features—such as convective storms, sea breezes, or orographic lift over terrain—can significantly influence local weather outcomes. A finer grid captures these features more accurately, enabling sharper forecasts of both typical conditions and extreme events. However, higher resolution entails greater computational demands and more intricate calibration needs, as small biases can propagate and amplify differently across the forecast horizon. The ongoing challenge for the forecasting community is to balance resolution, ensemble size, accuracy, and timeliness while maintaining robust probabilistic interpretation.
In practice, ensemble forecasts serve as decision-support tools rather than definitive predictions. They require careful interpretation, including model weighting, calibration, and post-processing steps that align ensemble outputs with observed realities. Properly designed, an AI-enhanced ensemble can improve both the sharpness of forecasts (the ability to predict precise outcomes) and the reliability (the degree to which forecast probabilities reflect actual frequencies) across a broad range of weather regimes, from routine days to severe-weather episodes.
GenCast is positioned within this landscape as a high-resolution AI-driven ensemble approach. By combining state-of-the-art machine learning techniques with an ensemble framework, GenCast aims to deliver more accurate and timely forecasts across diverse weather events. The model’s design emphasizes two core strengths: (1) the ability to resolve fine-scale atmospheric processes at a 0.25-degree grid, and (2) the integration of probabilistic predictions that quantify uncertainty in a way that is directly useful for risk-based decision-making. The claim is that GenCast outperforms the current top operational ensemble system up to 15 days ahead for both day-to-day weather and extremes, signaling a meaningful step forward in the practical utility of probabilistic forecasting.
To maximize impact, the GenCast project also emphasizes openness—making the code, the trained model weights, and the forecast outputs publicly available to the weather forecasting community. This approach aligns with broader trends toward reproducibility, independent validation, and collaborative improvement, all of which can accelerate the adoption of advanced AI-driven forecasting methods while enabling more robust performance assessments across different regions and climatologies.
GenCast: A High-Resolution AI Ensemble Model for Weather Forecasts
GenCast represents an ambitious effort to fuse cutting-edge artificial intelligence with the proven framework of ensemble weather prediction. The model operates at a high spatial resolution of 0.25 degrees, which translates to finer geographic detail and the potential to better resolve localized weather features. This granularity matters for forecast usefulness on regional scales where small differences in terrain, land use, and surface conditions can influence wind patterns, temperature distribution, and precipitation.
At its core, GenCast is an AI-enhanced ensemble system. It draws on a diverse set of initial conditions and model configurations to generate multiple plausible weather scenarios. Each member of the ensemble contributes to a richer representation of potential futures, especially under the influence of nonlinearity and chaotic dynamics that characterize atmospheric behavior. The ensemble probability distributions produced by GenCast furnish weather forecasters and end users with richer information than a single deterministic run could provide.
A distinctive claim of GenCast is its improved forecast skill for both routine weather patterns and extreme events when compared to the foremost operational ensemble system currently in routine use. The comparison highlights an improvement window extending up to 15 days into the future, a horizon that captures both near-term variability and the onset potential for significant deviations from average conditions. The assertion implies that GenCast achieves higher forecast accuracy and more reliable uncertainty estimates within this critical time frame, offering decision-makers a more trustworthy view of potential weather outcomes.
Beyond raw performance, GenCast emphasizes practical utility. The ensemble outputs are designed to be interpretable and actionable for a wide range of users, from meteorologists and researchers to public safety officials and industry stakeholders. The probabilistic forecasts can support risk communication by translating numerical confidence into clear statements about the likelihood of events such as heavy rainfall, heatwaves, or strong winds. In addition, possessing a 0.25-degree resolution can improve subregional assessments, enabling more effective situational awareness for communities, infrastructure planners, and emergency response teams.
A central feature of GenCast is the commitment to openness and community engagement. The project plans to release the model code, the trained weights, and the forecast products to the broader forecasting community. This openness is intended to foster collaborative validation, enable independent benchmarking across diverse geographic regions, and facilitate adaptation to local forecasting needs. By providing access to the underlying algorithms and data, the project aims to accelerate scientific scrutiny, reproducibility, and subsequent improvements across the field.
In terms of methodology, GenCast integrates AI techniques with an ensemble framework in a manner designed to enhance both forecast accuracy and the reliability of uncertainty estimates. While the exact technical details are beyond the scope of this summary, the approach typically involves training AI components on historical weather data, incorporating physical constraints, and calibrating outputs to ensure that predicted probabilities align with observed frequencies. The resulting forecasts are presented as probabilistic fields, enabling users to gauge not only the most likely outcomes but also the relative confidence associated with different weather scenarios.
The potential impact of GenCast spans multiple sectors. For the energy sector, more reliable forecasts can improve grid planning and renewable resource management. For agriculture, improved day-to-day predictions and better anticipation of extreme conditions can inform planting decisions, irrigation planning, and risk mitigation strategies. Transportation networks, including aviation and shipping, stand to benefit from clearer probabilistic guidance on weather-related disruptions. Public safety agencies can use enhanced forecasts to optimize resource allocation and emergency response planning ahead of hazardous weather events. In short, GenCast aims to translate advances in AI and high-resolution modeling into tangible improvements in forecast usefulness and decision support.
As part of its stated trajectory, GenCast emphasizes not only technical performance but also practical deployment and validation. The plan to share code, weights, and forecasts publicly is a deliberate move to encourage independent testing, cross-region comparisons, and real-world adoption. By enabling other researchers and forecasters to run, test, and refine GenCast, the project seeks to build a robust ecosystem in which AI-enhanced ensemble forecasting can mature through collaborative effort and transparent evaluation.
Benchmarking GenCast Against ECMWF’s ENS
The global forecasting community has long relied on established operational systems that combine advanced physics, data assimilation, and ensemble prediction methodologies. One of the most influential and widely used ensemble systems is ECMWF’s ensemble prediction system (ENS), which has served as a benchmark for many years due to its reliability, calibration, and broad utility across forecasting horizons. GenCast positions itself as a challenger to this benchmark by offering higher resolution and AI-driven enhancements that translate into improved forecast skill, particularly within the first two weeks of the forecast period.
The key claim, as reported, is that GenCast yields better forecasts than ECMWF’s ENS for both routine day-to-day weather conditions and extreme events, extending up to 15 days ahead. This implies gains in forecast accuracy, as well as more reliable probabilistic guidance over a horizon that is critical for planning, risk assessment, and early warning systems. The practical significance of such improvements is considerable: more accurate forecasts at these lead times can translate into reduced false alarms, better resource allocation, and more timely protective actions for populations and infrastructure.
Understanding how GenCast achieves these gains involves examining several aspects of model design and evaluation. High-resolution data assimilation, more effective representation of small-scale processes, and advanced calibration of ensemble probabilities are all relevant components. The integration of AI within an ensemble framework can offer improvements in pattern recognition, bias correction, and the generation of realistic ensemble members that span a wide range of weather scenarios. In benchmarking exercises, researchers typically evaluate performance across a variety of weather regimes, including temperature extremes, heavy precipitation events, wind storms, and cloud cover dynamics, to ensure that improvements are robust rather than limited to specific conditions.
An essential part of credible benchmarking is transparent, multi-faceted validation. This includes comparing forecast fields against observations across many regions and times, assessing reliability (the degree to which forecast probabilities match observed frequencies), sharpness (the concentration of forecast probabilities around the correct outcomes), and discrimination (the ability to distinguish between events that occur and those that do not). It also involves evaluating extreme-event predictability, where the costs of misprediction can be particularly high. While the exact figures and methods used in GenCast’s evaluation are not provided in this summary, the claim of superiority up to a 15-day horizon suggests consistent improvements in these key metrics across a substantial portion of the forecast space.
The broader implications of GenCast surpassing a leading ensemble system extend beyond technical performance. Forecasters may gain access to more nuanced input for decision-support tools, enabling more precise probabilistic statements like “there is a 40% chance of rainfall exceeding 25 mm in the next 48 hours” or “the likelihood of sustained winds above a critical threshold is 25% over the next week.” With improved reliability, public communications can become more credible, reducing the risk of message fatigue from false alarms or understatements of risk. Organizations involved in disaster preparedness, infrastructure planning, and climate resilience can leverage enhanced ensemble information to optimize readiness activities, allocate resources more efficiently, and refine contingency plans.
The decision to release model code, weights, and forecasts publicly reinforces the benchmarking framework by enabling independent replication and cross-validation. It invites the forecasting community to reproduce the results, test GenCast across different regions and climates, and integrate GenCast outputs into local forecasting workflows. Open access to forecasts also supports local scientific programs that need actionable probabilistic guidance to protect communities and maintain essential services during adverse weather events. The collaborative potential is significant: researchers can build on GenCast’s approach, compare it with other AI-augmented ensembles, and collectively push the frontiers of forecast accuracy and uncertainty quantification.
In evaluating GenCast relative to ECMWF ENS, it is important to acknowledge that both systems will have strengths and limitations across different weather regimes and geographies. An ensemble system’s performance can be regionally dependent, with certain climates and terrain features posing unique challenges. Even with improvements, the need for ongoing calibration, validation, and regional adaptation remains essential. The GenCast initiative—by aiming for open sharing and broad community engagement—addresses these considerations by inviting diverse, real-world testing and iterative refinement across a wide array of observational networks, forecast horizons, and end-user needs.
Implications for Forecasting Practice and Decision-Making
The introduction of a high-resolution AI ensemble like GenCast carries meaningful implications for forecasting practice and the way decisions are made in response to weather risk. First, the enhanced ability to forecast both routine conditions and extreme events with improved skill and well-calibrated uncertainty can lead to more nuanced risk assessments. For instance, energy grid operators may leverage higher confidence in temperature and wind forecasts to optimize generation mix and demand response strategies, while agricultural managers can time irrigation and harvest windows with greater precision when probabilistic forecasts reveal higher probabilities of heat stress or heavy rainfall.
Second, the 0.25-degree resolution offers more precise guidance in densely populated or geographically complex zones. When forecasts differentiate conditions at sub-regional scales, emergency managers can tailor evacuation plans, sheltering strategies, and resource pre-positioning to the needs of particular neighborhoods or districts. Transport networks—airports, railways, seaports, and highway corridors—benefit from improved confidence in weather-driven disruption risk, enabling smoother traffic management, flight scheduling, and maintenance planning. Public safety communications can rely on probabilistic statements to convey not only what is likely to occur but what the range of possible outcomes could be, thereby supporting risk-informed decision-making by individuals and organizations.
Third, the emphasis on ensemble probability and uncertainty quantification helps reduce the risk of overconfidence in any single forecast. Rather than presenting a single “best guess,” forecasters can provide a probabilistic outlook that communicates confidence levels, enabling stakeholders to weigh trade-offs and choose actions that align with their risk tolerance and operational constraints. In sectors like construction, energy, and logistics, this translates into more robust contingency planning, as decisions can be conditioned on a spectrum of plausible weather scenarios rather than a singular forecast.
Fourth, enabling the broader community to access GenCast’s code, weights, and forecasts accelerates the translation from research to practice. Open access supports capacity building in regions with limited forecasting resources, fosters local adaptation to regional climates, and invites university and industry partners to contribute to calibration and validation efforts. This democratization of advanced forecasting tools can enhance global preparedness for climate-related risks and support equitable resilience across diverse communities.
Fifth, the release strategy invites attention to the governance and ethics of AI-driven weather predictions. As AI components become more central to forecasting, stakeholders must consider issues such as data provenance, model transparency, bias mitigation, and the communication of uncertainty in a way that remains accessible to non-experts. Operational forecasters may require user-friendly interfaces and decision-support tools that translate probabilistic outputs into actionable guidance, while ensuring that the limitations and assumptions of GenCast are clearly understood by end users. The combination of AI sophistication and human-in-the-loop decision-making will shape the responsible deployment of such technology in real-world contexts.
Finally, the long horizon of up to 15 days challenges both forecasters and decision-makers to rethink planning cycles. For many operational contexts, planning horizons extend across days to weeks, and weather risk interacts with other temporal dynamics such as ecosystem responses, agricultural calendars, and infrastructure maintenance schedules. GenCast’s expanded horizon, coupled with higher resolution, enables more proactive planning—anticipating potential disruptions before they become acute and enabling time to mobilize resources, adjust operations, or implement mitigation measures.
Open Science and Community Impact
A notable dimension of GenCast is its commitment to openness. Making the model code, trained weights, and forecast products available to the broader forecasting community is a strategic choice aimed at accelerating scientific progress and practical adoption. Open access supports several important benefits:
- Reproducibility: Independent researchers can verify results, reproduce experiments, and assess robustness across different datasets and regions. Reproducibility is a cornerstone of credible scientific progress and a critical factor in building trust in AI-enabled forecasting systems.
- Benchmarking and cross-validation: With publicly available code and weights, the community can benchmark GenCast against other models, replicate comparative studies, and conduct standardized evaluations that ensure fair, apples-to-apples comparisons.
- Regional adaptation: Local forecasters can customize GenCast outputs to reflect regional climate characteristics, incorporating observational networks, terrain features, and land-use patterns specific to their area. This customization can improve local forecast skill and end-user relevance.
- Collaborative development: An open ecosystem invites contributions from researchers, meteorologists, software engineers, and domain experts, fostering innovation and iterative improvements. Community-driven enhancements can address edge cases, calibration challenges, and user-interface design to maximize practical impact.
- Education and capacity building: Access to cutting-edge forecasting tools supports training and skill development for students, early-career scientists, and practitioners in weather services around the world. This can broaden the pipeline of talent contributing to future advances.
At the same time, open releases require careful attention to documentation, user support, and governance. Adequate documentation helps users understand model inputs, outputs, performance characteristics, and limitations. User support channels and versioning practices help ensure that communities can adopt GenCast responsibly and with a clear understanding of changes over time. The open science approach thus represents not only a means to accelerate innovation but also a framework for shared responsibility in deploying advanced forecasting technologies.
The move toward openness also has implications for international collaboration in weather services and climate science. By providing accessible tools and outputs, GenCast can serve as a catalyst for joint validation studies, shared best practices, and collaborative research that considers diverse climatic contexts. This collaborative spirit aligns with the broader mission of improving global resilience to weather and climate risk through scientifically rigorous, widely accessible forecasting capabilities.
Potential Limitations and Challenges
While GenCast presents an exciting advancement, it is essential to acknowledge potential limitations and challenges that accompany any AI-assisted, high-resolution ensemble system. A balanced perspective helps practitioners set realistic expectations and design appropriate risk management strategies.
Data quality and representativeness: The performance of GenCast hinges on the quality and breadth of observational data used for training and calibration. Regions with sparse observational networks or inconsistent data coverage may pose challenges, potentially introducing regional biases or reduced forecast reliability. Addressing these gaps requires ongoing data assimilation improvements, sensor enhancements, and targeted regional validation campaigns.
Calibration and bias correction: AI-driven components often rely on calibration procedures to ensure that forecast probabilities reflect observed frequencies. If calibration drifts over time or under novel climate conditions, forecast reliability can be compromised. Continuous recalibration and robust evaluation across seasons and climate regimes are necessary to maintain trust in ensemble probabilities.
Interpretability and user communication: While probabilistic outputs offer rich information, communicating uncertainty to diverse end users remains a challenge. Forecast interpreters must develop clear, accessible, and actionable guidance that conveys both likely outcomes and associated uncertainties without overwhelming users with technical detail. User interfaces, visualization tools, and decision-support dashboards must be designed with cognitive load, regional relevance, and stakeholder needs in mind.
Computational demands: Maintaining 0.25-degree resolution within an AI-augmented ensemble requires substantial computational resources, including data storage, processing power, and energy consumption. Operational deployments must consider cost, efficiency, and sustainability while preserving forecast latency requirements that are critical for timely warnings and decision-making.
Model drift and climate change: As the climate evolves, atmospheric patterns and extremes may shift in ways that affect model performance. GenCast will need to be resilient to non-stationarity, with adaptive training strategies and ongoing monitoring to detect and correct for drift in forecast skill or calibration biases.
Regional and sector-specific validation: Different sectors (agriculture, energy, transportation, public safety) have unique needs and risk tolerances. Performance improvements observed on average across a broad domain may not translate uniformly to all sectors or regions. Sector-specific validation, experience-based calibration, and close collaboration with end users are essential to ensure forecasts meet practical requirements.
Privacy and data governance considerations: With open models and outputs, it is important to manage data governance securely, ensuring that sensitive information (where applicable) is handled responsibly and that data use adheres to relevant regulations and ethical standards.
Operational resilience and risk management: The deployment of AI-based forecasts introduces dependencies on software, hardware, and network infrastructure. Ensuring robust uptime, secure access, and failover capabilities is essential to maintain forecast availability during critical weather events.
In sum, GenCast represents a significant step forward, but its successful adoption hinges on ongoing validation, responsible deployment, and sustained collaboration across the forecasting community. Recognizing and addressing these limitations will help maximize the benefits of high-resolution AI ensemble forecasting while maintaining trust and reliability in weather risk assessments.
Future Directions and Research Pathways
Looking ahead, GenCast opens several avenues for further research and development that could broaden its capabilities and deepen its impact on weather forecasting and risk management.
-
Multi-model ensembles: Integrating GenCast with complementary models and observational streams could yield richer, more robust ensemble distributions. A multi-model ensemble approach can leverage diverse algorithmic perspectives and physical representations, potentially improving forecast reliability and reducing systematic biases.
-
Regional specialization: Tailoring GenCast to specific climatic zones and topographies could enhance local forecast skill. Region-specific calibrations, assimilation of high-density local observations, and regionally tuned post-processing can improve performance where it matters most to end users.
-
Probabilistic storytelling: Developing advanced visualization and interpretation tools to convey forecast probabilities and scenario narratives more intuitively could help non-expert audiences understand risk and make informed decisions. Narrative-based guidance alongside quantitative metrics can bridge gaps between meteorology and practical action.
-
Early warning optimization: Refining thresholds and lead times for hazard alerts based on probabilistic forecasts can improve the timing and effectiveness of warnings. This includes optimizing alert triggers to balance missed events against false alarms, ultimately protecting lives and property more efficiently.
-
Climate-informed forecasting: Extending GenCast to account for longer-term trends and climate variability could enhance predictability for climate-sensitive regimes. This might involve incorporating climate projections, seasonal forecasts, and ENSO-like signals into the ensemble framework.
-
Efficiency improvements: Research into more efficient neural architectures, data compression, and hardware accelerators can reduce computational costs without sacrificing accuracy. This would facilitate deployment across a wider range of forecast centers, including those with limited resources.
-
Federated validation and governance: Coordinating validation efforts across multiple forecast centers through standardized protocols can improve comparability and trust. Federated approaches can protect data sovereignty while enabling shared insights and performance benchmarks.
-
Education and outreach: Building curricula and training programs around AI-enhanced ensemble forecasting will prepare the next generation of meteorologists and data scientists to develop, validate, and responsibly deploy advanced tools.
-
Ethical and societal dimensions: Ongoing examination of the ethical implications, equity considerations, and communication strategies associated with AI-driven weather forecasts will help ensure that the benefits reach diverse communities while mitigating potential harms or misinterpretations.
GenCast’s trajectory toward open, collaborative development positions it well to evolve along these pathways, with continued validation, regional adaptation, and user-centered design driving improvements. By pursuing these directions, the forecasting community can build increasingly reliable and actionable weather predictions that support resilience in the face of a changing climate.
Global Impact, Accessibility, and Stakeholder Engagement
The potential global impact of GenCast stems from its improved accuracy, finer resolution, and transparent sharing model. Regions with advanced forecasting infrastructure stand to benefit from enhanced forecast skill and more reliable uncertainty quantification, enabling better risk management for weather-sensitive activities. At the same time, openness reduces the barriers to entry for countries and institutions with fewer resources, allowing them to leverage state-of-the-art forecasting tools without prohibitive licensing or development costs.
Stakeholders across sectors—government agencies, utilities, farmers, transport operators, insurers, and public safety organizations—stand to gain from more precise probabilistic forecasts. The ability to quantify and communicate risk with greater clarity can improve planning, emergency preparedness, and resilience-building efforts. For example, better anticipation of heavy precipitation or heat waves can inform water resource planning, crop protection measures, and heat-health warnings.
Education and capacity-building are natural beneficiaries of GenCast’s open approach. Universities and research institutes can study, replicate, and extend GenCast’s methods, training students to work at the intersection of meteorology and AI. This broad participation can accelerate skill development and cultivate a diverse community of practitioners with expertise in AI-augmented forecasting.
However, deepening the global reach of GenCast also requires attention to localization. Forecasts must be translated into regionally relevant guidance, taking into account local decision-making processes, cultural contexts, and language needs. User interfaces, documentation, and support services should be accessible to a global audience, embracing multilingual materials and region-specific case studies.
Collaboration with international weather networks, climate service providers, and end-user communities will be essential to maximize GenCast’s positive impact. Transparent, ongoing dialogue about model performance, limitations, and user needs will help ensure that the technology delivers tangible benefits where they are most needed, including in vulnerable communities that face disproportionate exposure to weather-related risks.
Conclusion
GenCast marks a significant milestone in the evolution of weather forecasting, representing a concerted effort to enhance both the accuracy of forecasts and the clarity of uncertainty at a critical lead time. By delivering high-resolution 0.25-degree ensemble forecasts, GenCast aims to improve the predictability of everyday weather and the forecasting of extreme events up to 15 days in advance. The model’s design leverages the strengths of probabilistic ensembles—providing a fuller picture of potential outcomes and their likelihoods—while leveraging advanced AI techniques to refine the representation of atmospheric processes and their interactions.
A central element of GenCast’s vision is openness: releasing the model code, trained weights, and forecast products to the broader community. This approach seeks to foster independent validation, cross-regional testing, and collaborative refinement, ultimately accelerating scientific progress and the practical adoption of AI-enhanced forecasting tools across diverse climates and regions. The promise of improved forecast skill, combined with more reliable uncertainty quantification, carries broad implications for decision-making across sectors that depend on weather information—energy, agriculture, transportation, emergency management, and public safety, among others.
GenCast’s performance claims—posited as outperforming the leading operational ensemble system for both routine conditions and extreme events within a 15-day horizon—signal a potential shift in how forecast guidance is generated, communicated, and used in practice. If realized through rigorous validation and careful implementation, these advances could translate into more informed risk management, better resource allocation, and enhanced resilience to weather-related disruptions. Yet, as with any AI-augmented forecasting system, careful attention to calibration, regional validation, proper interpretation of probabilistic outputs, and transparent communication with end users remains essential.
Looking forward, GenCast opens pathways to further research, testing, and refinement. The ongoing pursuit of higher resolution, more accurate probabilistic forecasts, and more effective end-user communication will help ensure that the benefits of this technology are realized broadly and responsibly. By continuing to engage with the forecasting community, end users, and stakeholders worldwide, GenCast can contribute to a future where weather information is not only faster and more precise but also clearer in its uncertainties, enabling smarter decisions and better resilience in the face of an increasingly variable climate.