SpaceXAI — Elon Musk's AI company, formerly known as xAI — released Grok 4.5 on July 8. Musk described it as an "Opus-class" model, placing it directly against Claude Opus from Anthropic, one of the most capable publicly available AI systems on the market.
Grok 4.5 runs on the V9 architecture with 1.5 trillion parameters — roughly three times the scale of Grok 4.3. The model was trained at the Colossus computing cluster in Memphis using Nvidia Blackwell GPUs. Developer data from Cursor was incorporated into supplemental training, giving the model a strong focus on software development tasks.
Pricing is the headline argument: $2 per million input tokens and $6 per million output tokens. That's considerably cheaper than rival Opus-tier models while claiming comparable performance. Grok 4.5 is available in Grok Build, across all Cursor subscription plans, and through the SpaceXAI API. EU users don't have access yet.
The timing is deliberate. OpenAI is reportedly preparing GPT-5.6, and SpaceXAI's pricing move puts immediate pressure on the market. If Grok 4.5 delivers on its benchmarks, competitors will need to justify why their own top-tier models cost significantly more.
