Allen AI has released OlmoEarth v1.1, a new family of transformer-based models designed for processing satellite imagery. These models achieve up to a 3x reduction in compute costs while maintaining the performance of the previous version. The efficiency gains are primarily realized by reducing the number of tokens processed, a key factor in transformer model computation costs, by merging resolution-based patches into single tokens without significant performance degradation. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Offers significant cost reductions for satellite imagery analysis, potentially enabling wider adoption and faster processing for environmental monitoring and mapping tasks.
RANK_REASON This is a release of a new version of an AI model family with a focus on efficiency improvements and technical details on how those were achieved, accompanied by a paper. [lever_c_demoted from research: ic=1 ai=1.0]