News
xAI Launches Grok 4.5, Priced Below Claude and GPT, but Blocked in the EU
xAI unveiled Grok 4.5, the first model built jointly with the newly acquired Cursor, undercutting Anthropic and OpenAI on coding costs. The model doesn't yet work in the European Union, including Poland, while the bloc completes its systemic risk assessment.
xAI announced on July 8 the launch of Grok 4.5, a new language model the company is marketing as an Opus-class tool, but faster and notably cheaper. It's the first model built since SpaceX's $60 billion acquisition of the startup Cursor, and xAI's first major release since its parent company's stock market debut.
Elon Musk described Grok 4.5 as comparable to Anthropic's Opus 4.7, but faster and cheaper per token. xAI says the new model was built with coding and agentic work in mind, scenarios in which the model carries out multi-step programming tasks on its own without constant human oversight.
Price as the Main Weapon
The price gap between Grok 4.5 and Anthropic's models is significant. On output tokens, Grok costs less than a quarter of Opus 4.7's rate, which for companies billing large volumes of agentic queries could meaningfully cut API bills. xAI speaks directly of twice the token efficiency compared to competitors, meaning fewer tokens are needed to complete the same task.
Benchmark comparisons show Grok 4.5 landing just behind the best models on the market, not topping any single major benchmark, but offering a noticeably lower cost at comparable quality. For many development teams, it's the price-to-performance ratio, not raw model power, that decides which tool becomes part of the daily workflow.
Cursor in the Background
The release comes weeks after SpaceX announced its $60 billion all-stock acquisition of Cursor, the popular coding agent. The deal is expected to close in the third quarter of this year, and Grok 4.5 is the first visible product of the two teams working together. The model is available by default in Cursor across all plans, putting it in direct competition with Claude and GPT-5.6 among developers who use that editor.
The combined effort also explains why xAI, a company known mainly for its Grok chatbot, is suddenly pushing a model specialized in coding. Earlier, the company had warned employees to limit contact with the Cursor team to avoid antitrust issues before the deal formally closed.
Blocked in Europe
Grok 4.5 doesn't yet work in the European Union, including Poland. The reason isn't a business decision by xAI, but the EU's AI Act. Because the model was trained above the compute threshold considered a systemic risk, xAI must pass mandatory safety testing, adversarial evaluations, incident reporting procedures and cybersecurity audits overseen by EU institutions before it can be offered in the bloc. Access for European users is expected by mid-July.
For Polish companies and developers, that means a few weeks of waiting, or using the model through foreign accounts, something providers explicitly advise against since it violates terms of service and risks account suspension. In practice, a short delay is cheaper than risking the loss of access to production deployments.
A War of Words
The Grok 4.5 launch coincided with a public exchange between Elon Musk and Sam Altman. On Saturday, Musk wrote on the X platform that Altman "takes cheating to a whole new level," without specifying the accusation. Altman responded by touting the performance of OpenAI's new model, without directly addressing the claim.
It is an Opus-class model, but faster, more token-efficient and lower cost - Elon Musk, xAI
elon is obsessed with me again - Sam Altman, OpenAI
The spat fits into a broader pattern of rivalry, including Musk's earlier, unsuccessful lawsuit against Altman over OpenAI's conversion from a non-profit into a for-profit company. Both companies released their flagship models almost simultaneously, with GPT-5.6 arriving a day after Grok 4.5, which alone fuels the marketing rivalry.
For the Polish market, though, what matters most is the practical side: when Grok 4.5 actually becomes available locally, and whether the lower price translates into a real reduction in costs for companies using coding agents. Until then, developers in Poland are sticking with Claude, GPT-5.6 and local alternatives.
Sources: TechCrunch (techcrunch.com), Axios (axios.com), x.ai (x.ai), Forbes (forbes.com), Trending Topics (trendingtopics.eu)


