The paper presents the Ocean-linked-atmosphere (Ola) model, a high-resolution (0.25°) Artificial Intelligence/ Machine Learning (AI/ML) coupled earth-system model designed to simulate ocean and atmosphere dynamics using an autoregressive Spherical Fourier Neural Operator (SFNO) architecture. The model aims to enable fast, accurate, and large ensemble forecasts on seasonal timescales, focusing on the El Niño/Southern Oscillation (ENSO) phenomenon. Key findings include: 1. **Seasonal ENSO Simulation**: Ola successfully generates realistic ENSO variability, including El Niño and La Niña states, with appropriate phase speeds and internal dynamics. The model captures the spatial and temporal characteristics of ENSO, such as the tropical Pacific SST anomalies and the vertical structure of ocean temperature anomalies. 2. **Performance Compared to Physics-Based Models**: Ola outperforms the GFDL-SPEAR model in terms of tropical drift bias, showing less systematic cold bias in the Niño 3.4 region. This suggests that machine learning models can avoid biases associated with subgrid parameterizations. 3. **Equatorial Oceanic Waves**: The model accurately simulates equatorial Kelvin and Rossby waves, with appropriate phase speeds and propagation directions, indicating successful coupling of ocean and atmosphere dynamics. 4. **Limitations and Future Work**: While Ola shows promise, it still exhibits long-term drifts at higher latitudes, which need further investigation. The model's performance is limited by the short evaluation period and the need for more extensive training data. Future work will focus on improving long-term stability and expanding the ocean state vector to better resolve oceanic physics. 5. **Methodology**: The Ola model is trained using the ERA5 reanalysis dataset for the atmosphere and the UFS-replay dataset for the ocean. The models are coupled using a specific interaction mechanism that links atmospheric and oceanic variables, with different timesteps for each component. 6. **Conclusion**: Ola represents a significant step forward in using AI/ML for seasonal climate forecasting, demonstrating the potential for accurate and efficient predictions of ENSO and other long-range climate phenomena. Further research is needed to refine the model and extend its capabilities to more complex climate scenarios.The paper presents the Ocean-linked-atmosphere (Ola) model, a high-resolution (0.25°) Artificial Intelligence/ Machine Learning (AI/ML) coupled earth-system model designed to simulate ocean and atmosphere dynamics using an autoregressive Spherical Fourier Neural Operator (SFNO) architecture. The model aims to enable fast, accurate, and large ensemble forecasts on seasonal timescales, focusing on the El Niño/Southern Oscillation (ENSO) phenomenon. Key findings include: 1. **Seasonal ENSO Simulation**: Ola successfully generates realistic ENSO variability, including El Niño and La Niña states, with appropriate phase speeds and internal dynamics. The model captures the spatial and temporal characteristics of ENSO, such as the tropical Pacific SST anomalies and the vertical structure of ocean temperature anomalies. 2. **Performance Compared to Physics-Based Models**: Ola outperforms the GFDL-SPEAR model in terms of tropical drift bias, showing less systematic cold bias in the Niño 3.4 region. This suggests that machine learning models can avoid biases associated with subgrid parameterizations. 3. **Equatorial Oceanic Waves**: The model accurately simulates equatorial Kelvin and Rossby waves, with appropriate phase speeds and propagation directions, indicating successful coupling of ocean and atmosphere dynamics. 4. **Limitations and Future Work**: While Ola shows promise, it still exhibits long-term drifts at higher latitudes, which need further investigation. The model's performance is limited by the short evaluation period and the need for more extensive training data. Future work will focus on improving long-term stability and expanding the ocean state vector to better resolve oceanic physics. 5. **Methodology**: The Ola model is trained using the ERA5 reanalysis dataset for the atmosphere and the UFS-replay dataset for the ocean. The models are coupled using a specific interaction mechanism that links atmospheric and oceanic variables, with different timesteps for each component. 6. **Conclusion**: Ola represents a significant step forward in using AI/ML for seasonal climate forecasting, demonstrating the potential for accurate and efficient predictions of ENSO and other long-range climate phenomena. Further research is needed to refine the model and extend its capabilities to more complex climate scenarios.