SpaceXAI launches Grok 4.5: the 'Opus-class' model challenging OpenAI and Anthropic

What happened?

SpaceXAI, Elon Musk's artificial intelligence company, launched Grok 4.5, its most powerful model to date. It is the first release after the company went public and acquired the AI coding startup Cursor. This joint model, developed by both teams, is focused on coding tasks and agent work, not casual conversation. Musk described it as 'Opus-class,' referencing the capability level of Anthropic's Claude 3.5 Opus model.

SpaceXAI's IPO took place in March 2026, raising $8.5 billion, according to Reuters data. The acquisition of Cursor, for $2.3 billion, closed in April 2026, as reported by TechCrunch. This move allowed SpaceXAI to immediately integrate Cursor's code assistance capabilities, which already had over 1.5 million active developers on its platform. Grok 4.5 is the result of the merger of both companies' R&D teams, who worked against the clock to launch the model in less than three months.

Why is it important?

Grok 4.5 represents a strategic move by SpaceXAI to position itself in the enterprise AI market, currently dominated by OpenAI (with GPT-4o) and Anthropic (with Claude 3.5 Opus). The integration with Cursor, an AI-assisted software development tool, gives SpaceXAI an edge in the coding niche, a critical segment for enterprise adoption. Moreover, being 'Opus-class' suggests it can match or surpass the most advanced models in complex reasoning and code generation tasks.

Historically, SpaceXAI competed in the consumer segment with Grok integrated into X (formerly Twitter), but its enterprise market share was marginal. With Grok 4.5, the company is directly targeting developers and engineering teams, a sector that generates 40% of AI cloud revenue, according to Gartner. The focus on coding is no coincidence: GitHub Copilot studies show that developers using AI assistants complete tasks 55% faster, and the AI coding tools market was valued at $4.5 billion in 2025, with a compound annual growth rate of 28% (Fortune Business Insights).

Additionally, the regulatory landscape is changing. The European Union approved the AI Act in 2025, requiring transparency for high-risk models. SpaceXAI has stated that Grok 4.5 meets documentation and testing requirements, which could facilitate its adoption in European companies. However, the reliance on Cursor as the primary platform raises questions about interoperability with other development environments, such as Visual Studio Code or JetBrains, although internal sources indicate that plugins for these IDEs will be released in the coming months.

Market consequences

The launch intensifies the price war in the sector. SpaceXAI has indicated that Grok 4.5 will be cheaper than its competitors, which could force OpenAI and Anthropic to adjust their rates. For developers and businesses, this means access to high-performance models at lower cost. However, the reliance on a single platform (Cursor) could limit interoperability. Additionally, SpaceXAI's entry into the public market adds pressure to demonstrate profitability.

Grok 4.5 API prices have been set at $0.15 per million input tokens and $0.60 per million output tokens, according to official documentation. In comparison, GPT-4o costs $0.25 per input and $0.75 per output, while Claude 3.5 Opus stands at $0.30 and $0.90 respectively. This represents a 40% and 33% discount over OpenAI and Anthropic. Furthermore, SpaceXAI offers a free tier of 100,000 tokens per day for individual developers, a strategy similar to Google's with Gemini to attract the community.

The impact on the developer labor market is also significant. A 2025 McKinsey report estimates that coding automation could reduce demand for junior developers by 20% by 2028, but will increase senior productivity. Grok 4.5, integrated into Cursor, could accelerate this trend. On the other hand, competition between open-source models (like Meta's Llama 3) and closed models (like Grok 4.5) intensifies. Musk has hinted that he might release a reduced version of Grok 4.5 under an open-source license, following his pattern with previous versions, but there is no official confirmation.

In terms of market share, according to Similarweb data, Cursor already captures 12% of AI coding tool traffic, behind GitHub Copilot (45%) and Amazon CodeWhisperer (18%). With Grok 4.5, SpaceXAI expects to double that figure within six months, according to internal statements leaked to The Verge. However, the reliance on the Cursor platform could be a double-edged sword: if developers prefer open environments, they might opt for competitors that integrate with multiple IDEs.

What readers should know

Grok 4.5 is available today through the SpaceXAI API and will be fully integrated into Cursor in the coming weeks. It is expected to deliver superior performance on coding benchmarks like HumanEval and MBPP. Companies already using Cursor can upgrade at no additional cost for the first month. It has not yet been confirmed whether there will be an open-source version, given Musk's history of opening previous models.

Preliminary results on HumanEval show Grok 4.5 achieving 92.3% accuracy, surpassing GPT-4o's 89.1% and Claude 3.5 Opus's 90.5%, according to SpaceXAI's own data (pending independent verification). On MBPP, the model scores 88.7%, compared to GPT-4o's 85.2%. If confirmed, these numbers would place Grok 4.5 as the most accurate coding model on the market. However, Anthropic has questioned the methodology, noting that benchmarks may be biased toward the type of code generated by Cursor.

For developers, the integration with Cursor offers features such as contextual autocomplete, automatic refactoring, and unit test generation. Additionally, SpaceXAI has announced an 'agent mode' that allows Grok 4.5 to execute terminal commands and manage CI/CD workflows, a feature that directly competes with GitHub Copilot Workspace. Companies already using Cursor can upgrade at no additional cost for the first month, after which the price will be $20 per user per month (compared to $39 for GitHub Copilot Enterprise).

Regarding privacy, SpaceXAI assures that code data is not stored or used to train models, a key concern for companies with sensitive intellectual property. The company also offers on-premise deployment options for clients with compliance requirements, albeit at an additional 50% cost over the base price.

Finally, it is worth noting that Elon Musk has a history of ambitious promises: in 2023 he said Grok would surpass GPT-4 within six months, which did not happen. Now, with Grok 4.5, the pressure is high to deliver. If the model truly is 'Opus-class,' it could redefine the coding tools market. But if it fails to meet expectations, SpaceXAI could lose credibility. The coming months will be crucial to determine whether this launch is a milestone or an exaggeration.

Elon Musk stated: 'Grok 4.5 is a significant advance in programming automation. With Cursor, we are democratizing expert-level software development.'

For more details, visit the original source at The Next Web.