SpaceXAI Unveils Grok 4.5 as Elon Musk Calls It an Opus-Class AI Model

SpaceXAI has officially introduced Grok 4.5, its newest artificial intelligence model and the company’s first major release since it went public several weeks ago.

In a blog post published on Wednesday, SpaceXAI described Grok 4.5 as a general-purpose AI model capable of handling many of the tasks that have become central to the AI industry. According to the company, the model is designed for coding, application development, office and administrative work, research, writing, and a broad range of routine knowledge-based tasks.

SpaceXAI also claims that Grok 4.5 delivers these capabilities more efficiently than competing models. The company says the new model offers “twice the token efficiency” compared with other leading AI systems. If those gains are reflected in real-world deployments, improved token efficiency could become a significant advantage, particularly as the cost of AI inference continues to be a growing concern for businesses and developers.

Alongside the announcement, SpaceXAI published benchmark results that it says demonstrate Grok 4.5’s competitiveness against other leading foundation models. While the figures suggest the model performs at a high level, the published results stop short of claiming industry-leading performance across every benchmark.

Founder Elon Musk also promoted the launch on X, the social media platform owned by SpaceX AI, comparing Grok 4.5 with Anthropic’s Opus family of large language models, which are designed to handle demanding and complex workloads.

In his post, Musk said that, based on highly positive customer feedback from the company’s beta testing programme, SpaceX AI would make Grok 4.5 publicly available the following day. He described the model as “an Opus-class model,” while adding that it delivers faster performance, improved token efficiency, and lower operating costs.

Musk later expanded on that comparison, stating that the company’s internal evaluations suggest Grok 4.5 performs at a level roughly comparable to Anthropic’s Opus 4.7 while operating considerably faster. According to him, the combination of strong capabilities, higher speed, and lower pricing is what makes the model competitive in today’s AI market.

SpaceXAI has priced Grok 4.5 at $2 per million input tokens and $6 per million output tokens. If the model performs in line with the company’s claims, those prices would place it among the more competitively priced premium AI offerings currently available.

For comparison, Anthropic’s Opus 4.7 is priced at $5 per million input tokens and $25 per million output tokens. OpenAI also uses tiered pricing across its product lineup, with its high-end model, Sol, costing $5 per million input tokens and $30 per million output tokens. In comparison, its lower-priced Luna model is available at $1 per million input tokens and $6 per million output tokens.

The release comes during an especially active period for the AI industry. OpenAI is also preparing to introduce GPT-5.6, its newest and most advanced foundation model, on Thursday. Reports have indicated that the rollout of GPT-5.6 was previously delayed due to restrictions imposed by the Trump administration over security concerns. OpenAI has described GPT-5.6 as its most capable model to date.