July 8 - SpaceXAI released Grok 4.5 on Wednesday, presenting the model as the company’s most advanced offering to date for coding-oriented and agentic tasks. The company said the training process leveraged tens of thousands of Nvidia GB300 graphics processing units and placed particular emphasis on rigorous data filtering, deduplication and quality scoring.
SpaceXAI highlighted immediate access paths for developers and coding agents. Grok 4.5 can be used today through SpaceXAI’s coding assistant Grok Build inside the Cursor environment and through the SpaceXAI console, where developers can connect via an API key. SpaceXAI added that availability in the European Union is expected in mid-July.
On pricing, SpaceXAI set Grok 4.5 at $2 per million input tokens and $6 per million output tokens. The company framed the release in comparative terms: SpaceX CEO Elon Musk characterized the model as "an Opus-class model, but faster, more token-efficient and lower cost" in a post on X.
Industry context was provided within the company’s announcement. SpaceXAI cited a partnership with Cursor in training Grok 4.5; Cursor, the AI coding agent, said, "We’ve partnered with SpaceXAI to train Grok 4.5." The parent-company move that supports that relationship was noted: SpaceX said last month it would acquire Anysphere, the startup behind Cursor, in an all-stock transaction valued at $60 billion aimed at expanding SpaceX’s footprint in the enterprise AI tools market.
The release also referenced competing model economics. SpaceXAI’s posted prices for Grok 4.5 were contrasted with Anthropic’s Claude Opus 4.8, which is priced at $5 per million input tokens and $25 per million output tokens, and with OpenAI’s GPT-5.6 Luna, which is listed at $1 per million input tokens and $6 per million output tokens.
The article also noted recent corporate-level changes at Musk’s AI efforts: xAI was acquired by SpaceX in February, and in May Musk indicated xAI would no longer exist as a separate company and would instead become SpaceXAI.
Separately, the report referenced OpenAI’s upcoming public launch of GPT-5.6, scheduled for Thursday, following a prior delay attributed to requests from U.S. government officials regarding national security concerns about powerful AI technologies. The article explained the technical distinction between input and output tokens: input tokens are the text, code or other data sent to a model, while output tokens are the text or code the model generates in response.
- Availability: Grok 4.5 is available now through Grok Build in Cursor and the SpaceXAI console; EU rollout expected mid-July.
- Training infrastructure: Tens of thousands of Nvidia GB300 GPUs were used, with emphasis on data filtering, deduplication and quality scoring.
- Pricing: $2 per million input tokens and $6 per million output tokens.