MMAI-Trained Foundation Models Claim 10× Gains on Drug Discovery Benchmarks

A new training framework for multimodal AI shows that domain-specific foundation models outperform general-purpose models by an order of magnitude on key drug discovery tasks. General LLMs fell short on approximately half the tested benchmarks, reinforcing that biology needs purpose-built models.