Google AI, in collaboration with Google DeepMind, has introduced Tx-LLM, a new large language model fine-tuned from PaLM-2, designed to predict properties of entities relevant to therapeutic development. This model leverages data from 66 drug discovery tasks and has shown to outperform or match specialist models on 65% of these tasks. According to a press release, Tx-LLM is capable of predicting properties for tasks ranging from early-stage target identification to late-stage clinical trial approval. The model aims to streamline the drug development process, which is often time-consuming and complex, with accuracy comparable to state-of-the-art specialty models.