Earth-2 (cBottle-SR)

Training compute

1.6×10²¹ FLOP

Parameters

330M

Published

May 10, 2025

Earth-2 (cBottle-SR) is an AI model developed by NVIDIA (United States), first published in May 2025. It works in the earth science and image generation domain, on tasks such as image generation and weather forecasting.

Training it took an estimated 1.6×10²¹ FLOP of compute (estimation method: hardware). The model has 330,000,000 parameters. Training ran on 64 NVIDIA H100 SXM5 80GB.

Access: Open weights (non-commercial). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: NVIDIA
Country of organization: United States
Domain: Earth science, Image generation
Task: Image generation, Weather forecasting
Training compute: 1.6×10²¹ FLOP
Compute estimation method: Hardware
Parameters: 330,000,000
Training hardware: NVIDIA H100 SXM5 80GB
Chips used: 64
Chip-hours: 1.5K
Training power draw: 87.8 kW
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Epoch confidence: Confident