xAI unveiled the new AI model Grok 4.5 for programming and complex tasks

xAI has introduced a new language model, Grok 4.5, calling it the most powerful in its lineup. The model is designed for programming, agent-based scenarios, and solving complex intellectual tasks. Grok 4.5 was trained in collaboration with the Cursor development environment team.

The model demonstrates high performance in engineering tests and software development tasks, and is capable of creating complete applications from a single text prompt. Grok 4.5 supports working with languages such as Rust, C/C++, and others, and can automatically generate documents, presentations, spreadsheets, and charts.

The model was trained on tens of thousands of NVIDIA GB300 graphics processing units using large-scale reinforcement learning and specialized datasets for programming, science, engineering, and mathematics. According to xAI, Grok 4.5 operates at speeds of up to 80 tokens per second and requires approximately half the number of tokens to solve complex tasks compared to several competing models.

The API costs $2 per million input tokens and $6 per million output tokens. Grok 4.5 is already available in the Grok Build service, the Cursor environment, and via the xAI API.

Source: x.ai