Grok 4.5 launched; Musk claims it's faster and cheaper than Opus-level models

xAI, the artificial intelligence venture led by Tesla CEO Elon Musk, has officially launched its latest AI model, Grok 4.5. Developed in collaboration with the coding company Cursor, Grok 4.5 is designed to be particularly adept at coding, agent tasks, and knowledge-intensive work.

The company announced the release on May 8th, highlighting the advanced hardware used in its training. Grok 4.5 was trained using tens of thousands of Nvidia's latest GB300 graphics processing units. To ensure high-quality output, xAI also implemented a more rigorous data quality assessment process.

In benchmark tests comparing coding capabilities in a terminal environment, Grok 4.5 reportedly scored lower than OpenAI's GPT-5.5 model but higher than Claude's Opus 4.8 model. When evaluating coding performance in languages like Java and Python, Grok 4.5 surpassed GPT-5.5 but fell short of Opus 4.8.

A key selling point for Grok 4.5 is its cost-effectiveness. The model is priced between $2 to $6 per 1 million tokens, making it two to four times cheaper than Claude Opus 4.8, which ranges from $5 to $25 per 1 million tokens. Elon Musk emphasized this advantage on X, formerly Twitter, stating that Grok 4.5 offers "Opus-level performance but is faster, more token-efficient, and cheaper."

This launch follows xAI's acquisition in February and subsequent rebranding. The company has been actively integrating its capabilities with Musk's other ventures. In April, SpaceX announced its intention to acquire Anysphere, the operator of Cursor, in an all-stock deal valued at $60 billion, signaling a deepening collaboration between the teams on Grok's development.

“

Grok 4.5 is an Opus-level model, but faster, more token-efficient, and cheaper.