The Arc Institute has launched the Virtual Cell Challenge, a competition incentivizing development of AI models that predict gene expression responses to genetic perturbations across diverse cellular contexts. Competitors utilize extensive single-cell transcriptome datasets comprising over 300,000 human embryonic stem cells with 300 gene modifications. Sponsored by Nvidia, 10x Genomics, and Ultima Genomics, the initiative emphasizes generalization capability through few-shot learning approaches, aiming to benchmark and accelerate AI progress in systems biology and drug discovery.