Just as artificial intelligence has learned the nuances of human language, a new wave of research is focused on teaching these models to understand the chemical language of life. The challenge, however, is that our evaluation tools for these genomic models are often too simple, unable to fully capture the true complexity of DNA and RNA. To address this, we launched the OmniGenomic Benchmark (OGB) competition. The OGB served as a rigorous test for models, presenting a diverse suite of seven tasks designed to mimic the multifaceted nature of real-world genomic problems. By attracting researchers from around the globe, the competition showcased promising strategies for optimising gFMs in task-specific settings, and also demonstrated that advanced models can perform consistently across a wide range of genomic challenges. Ultimately, this competition marks an important milestone, establishing a new baseline for future work and highlighting the urgent need for a deeper collaboration between computer scientists and biologists to create models that can truly integrate the multi-modal data of life.