SpaceXAI's Grok 4.5 targets the coding stack with Opus-class claims at 60–75% lower prices, and the Cursor acquisition gives it a distribution flywheel rivals can't easily copy.
SpaceXAI released Grok 4.5 on July 8, its first model launch since going public and since agreeing to acquire Cursor parent Anysphere in a reported $60 billion all-stock deal, the largest startup acquisition on record. Elon Musk describes Grok 4.5 as "an Opus-class model, but faster, more token-efficient and lower cost."
The positioning is unambiguous: this is a coding and agentic-work model rather than a consumer chatbot. Grok 4.5 was trained alongside Cursor and is available in Grok Build, across all Cursor plans, and via the SpaceXAI console. EU availability is expected mid-July.
Pricing is the sharpest edge. At $2 per million input tokens and $6 per million output tokens, Grok 4.5 undercuts Anthropic's Opus 4.8 ($5/$25) by roughly 60–75%, while claiming comparable capability on SpaceXAI's own benchmarks. Independent evals will be the arbiter: on the third-party Vals Index, Grok 4.5 debuted at 65.30%, a near 19-point jump over Grok 4.3.
Our Take
the Cursor acquisition is already earning its keep as a distribution and training asset. Owning the IDE through which millions of developers consume models gives SpaceXAI both a proprietary data flywheel and a captive deployment surface, a vertical-integration play the other labs cannot easily replicate. The strategic risk sits with Cursor's historical position as a neutral multi-model front end; if enterprise users perceive the product steering toward Grok, competitors' models inside Cursor become a churn vector worth watching.