Elon Musk's SpaceXAI released its new artificial intelligence model, Grok 4.5, on Wednesday, its first public model since the SpaceX-xAI merger closed in February and its pending acquisition of coding startup Cursor. The company is positioning the model as a coding and agentic-work tool built for cost-conscious enterprise users rather than as a benchmark-topping frontier system.
Musk described Grok 4.5 in a post on X as an "Opus-class model," a reference to Anthropic's premium model tier, but claimed it was faster, more token-efficient and cheaper to run. He later clarified that the company's internal assessment placed it "roughly comparable" to Anthropic's Opus 4.7 in capability, while running considerably faster. Musk framed the pitch around real-world usefulness rather than leaderboard rankings, saying engineers at Tesla and SpaceX had found the model genuinely useful in practice.
SpaceXAI's central argument is economic rather than purely technical. Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens, undercutting Anthropic's Opus 4.8, which is priced at $5 and $25 respectively, and OpenAI's newly launched GPT-5.6 Sol, priced at $5 and $30. The company says the model uses roughly half the tokens per task compared with peer systems while delivering higher throughput, running at about 80 tokens per second.
Grok 4.5 was trained with Cursor using real-world coding data on SpaceXAI’s Colossus supercomputer and is now available through Grok Build, Cursor and the SpaceXAI console, excluding the EU.
In benchmark tests, Grok 4.5 delivered mixed results. It outperformed GPT-5.5 on SWE-Bench Pro but lagged behind Anthropic’s Claude models on both SWE-Bench Pro and DeepSWE 1.1.
The launch marks a key milestone for Musk’s AI venture after months of controversy. Grok previously faced backlash over antisemitic responses and AI-generated deepfakes, drawing regulatory attention in Europe and the UK. Following the SpaceX–xAI merger and a strong post-IPO performance, the company plans to roll out a new foundation AI model every month through the end of 2026 to strengthen its position against OpenAI and Anthropic.
