SpaceXAI announces the launch of Grok 4.5, positioning the model for coding and “agentic” tasks, according to the available reporting. The Hindu states that SpaceXAI says Grok 4.5 is trained across tens of thousands of Nvidia GB300 graphics processing units. Channel NewsAsia carries the same headline and topic but provides no additional details in the supplied excerpt. Across the sources, the central point is that Grok 4.5 is introduced as a newer Grok-family model from SpaceXAI, with a focus on developer-oriented use cases such as writing or assisting with code and performing tasks that can involve multi-step or autonomous workflows. The reports also emphasize the scale of the compute used for training, citing very large GPU resources and specific Nvidia hardware. No further information is provided in the excerpts on performance results, availability, pricing, or release timelines, and the sources do not indicate any competing claims or disagreements.