SpaceXAI (formerly xAI) has introduced a new Grok 4.5 model for programming and artificial intelligence agents. The company said that the model was developed in conjunction with the Cursor team. Recall that SpaceXAI announced the acquisition of the Cursor platform in June 2026.
According to the company, Grok 4.5 was trained on a large database covering programming, science and mathematics. While Cursor's Composer 2.5, introduced in May 2026, specializes mainly in programming, Grok 4.5 is designed for more universal tasks.
The model is built on the Mixture of Experts (MoE) architecture and shows high performance in performing complex, long-term tasks. According to the company, the model can be used in finance, law and other fields in addition to software development.
Grok 4.5 outperformed Anthropic's Opus 4.8 and Cursor's Composer 2.5 in the Terminal-bench 2.1 test, which evaluates AI agents.
Cursor CEO Michael Truell described the model as "Opus-level, fast, and low-cost." Grok 4.5 is currently available on Cursor, Grok Build, and SpaceXAI Console platforms. The cost to use the model is $2 per million input tokens and $6 per million output tokens.
