xAI has unveiled Grok 4.5, its latest large language model, which is now undergoing private beta testing. The model is being used exclusively by teams at SpaceX and Tesla.
The new model is built on xAI's V9 foundation, which contains 1.5 trillion parameters. This makes it roughly three times larger than the architecture that powered earlier Grok 4 versions.
The training pipeline was enhanced with data from the AI-powered code editor Cursor to improve coding and technical abilities. Early internal evaluations indicate performance comparable to, or better than, Anthropic's Claude Opus.
xAI has not announced a public release date. The model is already running about a month behind its original late May 2026 timeline. The company's aggressive roadmap calls for monthly model releases in 2026, with future Grok 5 variants projected to reach 10 trillion parameters.