SpaceXAI rolled out its latest artificial intelligence model on Wednesday, introducing Grok 4.5 as the company’s most advanced product to date, built specifically for coding and so-called agentic tasks — work that AI can carry out with greater independence.
The company said Grok 4.5 was developed using tens of thousands of Nvidia GB300 graphics processing units, with heavy emphasis on careful data filtering, removing duplicate information, and evaluating the quality of training data.
The popular AI coding tool Cursor confirmed its involvement in the project, stating, “We’ve partnered with SpaceXAI to train Grok 4.5.” SpaceX announced last month that it plans to acquire Anysphere, the startup that created Cursor, through an all-stock deal valued at $60 billion, a move aimed at strengthening its foothold in the enterprise AI tools market.
Grok 4.5 is now accessible through SpaceXAI’s own AI coding agent called Grok Build, within Cursor, and via the SpaceXAI console — the company’s developer portal — using an API key. Availability in the European Union is expected to follow in mid-July.
In terms of pricing, SpaceXAI set Grok 4.5 at $2 per million input tokens and $6 per million output tokens. For context, input tokens refer to the text, code, or other data a user sends to an AI model, while output tokens are what the model produces in return.
SpaceX CEO Elon Musk weighed in on the release through a post on X, describing the model as “an Opus-class model, but faster, more token-efficient and lower cost.”
The launch comes after Musk’s AI startup xAI was acquired by SpaceX back in February. Musk announced in May that xAI would no longer operate as a standalone company and would be folded into what is now called SpaceXAI.
By comparison, rival Anthropic’s Claude Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens. OpenAI’s GPT-5.6 Luna, meanwhile, is priced at $1 per million input tokens and $6 per million output tokens.
OpenAI is set to publicly release its most powerful AI model, GPT-5.6, on Thursday. That launch had been delayed last month after the U.S. government raised national security concerns about the possible misuse of highly capable AI technologies.
