This paper presents the Ocean-linked-atmosphere (Ola) model, a high-resolution (0.25°) artificial intelligence/machine learning (AI/ML) coupled earth-system model that separately models ocean and atmosphere dynamics using an autoregressive spherical Fourier neural operator (SFNO) architecture. The model is designed to enable fast, accurate, large ensemble forecasts on seasonal timescales. Ola exhibits learned characteristics of ocean-atmosphere coupled dynamics, including tropical oceanic waves with appropriate phase speeds and an internally generated El Niño/Southern Oscillation (ENSO) with realistic amplitude, geographic structure, and vertical structure within the ocean mixed layer. Initial evidence suggests that Ola's ENSO forecasts compare favorably to the SPEAR model of the Geophysical Fluid Dynamics Laboratory. The Ola model is able to capture characteristic ocean-atmosphere interactions that strongly influence seasonal outlooks of weather. The model's ability to simulate ENSO, which has a large impact on global weather and is a key source of long-range predictability, demonstrates a pathway towards accurate forecasts of long-range seasonal climate and uncertainty using AI models. The model's ocean component represents conditions spanning the Eastern Pacific thermocline, from the surface down to 300 meters. These ocean dynamics are conditioned on atmospheric winds, pressure, and temperature, and are linked to an SFNO atmosphere model, which is in turn conditioned on sea surface temperature. The model's results show that Ola generates realistic amounts of Central Tropical Pacific SST variability in the 5-year validation period. Ola forecasts can internally generate both El Niño and La Niña states from a neutral state and return to a neutral state when initialized in an El Niño or La Niña state. However, an exception occurred after two consecutive La Niña events when Ola predicted a shift from La Niña to neutral, whereas La Niña persisted in reality. The model also produces less of a time-mean cold bias in the Niño 3.4 region compared to the GFDL-SPEAR model, which is a well-documented issue in many process-based coupled ocean-atmosphere models. The model's results also show that Ola generates oceanic equatorial Kelvin and Rossby waves, which are critical to ENSO. The model's simulations of equatorial Kelvin and Rossby waves show basin crossing timescales of approximately 2 months, consistent with observations and theory. The model's simulations of ENSO ocean temperature anomalies show realistic three-dimensional thermal structure, with SST anomalies resembling observations and potential temperature anomalies skillfully resembling observations. The model's results also show that Ola's simulated ENSO events exhibit realistic three-dimensional thermal structure, with SST anomalies resembling observations and potential temperature anomalies skillfully resembling observations. The model's results also show that Ola's simulated ENSO events exhibit realistic three-dimensional thermal structure, with SST anomalies resembling observations and potential temperature anomalies skillfully resembling observations. The modelThis paper presents the Ocean-linked-atmosphere (Ola) model, a high-resolution (0.25°) artificial intelligence/machine learning (AI/ML) coupled earth-system model that separately models ocean and atmosphere dynamics using an autoregressive spherical Fourier neural operator (SFNO) architecture. The model is designed to enable fast, accurate, large ensemble forecasts on seasonal timescales. Ola exhibits learned characteristics of ocean-atmosphere coupled dynamics, including tropical oceanic waves with appropriate phase speeds and an internally generated El Niño/Southern Oscillation (ENSO) with realistic amplitude, geographic structure, and vertical structure within the ocean mixed layer. Initial evidence suggests that Ola's ENSO forecasts compare favorably to the SPEAR model of the Geophysical Fluid Dynamics Laboratory. The Ola model is able to capture characteristic ocean-atmosphere interactions that strongly influence seasonal outlooks of weather. The model's ability to simulate ENSO, which has a large impact on global weather and is a key source of long-range predictability, demonstrates a pathway towards accurate forecasts of long-range seasonal climate and uncertainty using AI models. The model's ocean component represents conditions spanning the Eastern Pacific thermocline, from the surface down to 300 meters. These ocean dynamics are conditioned on atmospheric winds, pressure, and temperature, and are linked to an SFNO atmosphere model, which is in turn conditioned on sea surface temperature. The model's results show that Ola generates realistic amounts of Central Tropical Pacific SST variability in the 5-year validation period. Ola forecasts can internally generate both El Niño and La Niña states from a neutral state and return to a neutral state when initialized in an El Niño or La Niña state. However, an exception occurred after two consecutive La Niña events when Ola predicted a shift from La Niña to neutral, whereas La Niña persisted in reality. The model also produces less of a time-mean cold bias in the Niño 3.4 region compared to the GFDL-SPEAR model, which is a well-documented issue in many process-based coupled ocean-atmosphere models. The model's results also show that Ola generates oceanic equatorial Kelvin and Rossby waves, which are critical to ENSO. The model's simulations of equatorial Kelvin and Rossby waves show basin crossing timescales of approximately 2 months, consistent with observations and theory. The model's simulations of ENSO ocean temperature anomalies show realistic three-dimensional thermal structure, with SST anomalies resembling observations and potential temperature anomalies skillfully resembling observations. The model's results also show that Ola's simulated ENSO events exhibit realistic three-dimensional thermal structure, with SST anomalies resembling observations and potential temperature anomalies skillfully resembling observations. The model's results also show that Ola's simulated ENSO events exhibit realistic three-dimensional thermal structure, with SST anomalies resembling observations and potential temperature anomalies skillfully resembling observations. The model