Grok 4.5 is xAI’s first model since the company went public, and Elon Musk described it as an “Opus class” model that targets coding, application development, and autonomous multi-step tasks.
Grok 4.5 focuses on coding and agent tasks
xAI said in a company blog post that Grok 4.5 is positioned as a workhorse for routine knowledge work, covering coding, app development, administrative writing, research, and creative writing.
The company emphasized the model’s ability to run agentic tasks, meaning it can carry out multi-step processes without human intervention, such as combining web research to build a complex Excel financial model or producing a well-structured PowerPoint slide deck.
Training, Cursor data, and infrastructure
According to xAI’s announcement, Grok 4.5 was trained on tens of thousands of Nvidia GB300 GPUs and incorporates coding data from Cursor, the developer tool xAI acquired earlier this year.
The company said the training effort focused on data filtering, deduplication, and quality scoring, and that internal evaluations place Grok 4.5’s capabilities close to Opus 4.7 while being noticeably faster.
Pricing undercuts Opus by more than 60 percent
xAI published per-token pricing that it says is competitive with leading models. Grok 4.5 charges approximately $1.99 per million input tokens (originally HK$15.6) and approximately $5.97 per million output tokens (originally HK$46.8), according to the company blog post.
For comparison, Anthropic’s Claude Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens, while OpenAI’s GPT-5.6 Luna is listed at $1 per million input tokens and $6 per million output tokens.
By xAI’s numbers, Grok 4.5’s output cost is roughly 24 percent of Opus 4.8’s output cost, a meaningful difference for developers running large numbers of agent tasks.
Availability and access
Grok 4.5 is available immediately through Grok Build, the Cursor platform for all plans, and the SpaceXAI developer console, with API key access for integrations, xAI said.
xAI expects the model to be available in the European Union around mid-July, the company said, without providing a specific date for broader rollout.
Industry takeaways
xAI has pitched Grok 4.5 as an enterprise tool rather than a consumer chat assistant, reflecting a shift in the market from benchmark performance to cost efficiency and task usefulness, according to the company blog post and reporting by TechCrunch and Axios.
Observers say Grok 4.5’s lower token prices and agent capabilities make it an attractive option for developers building automated workflows, but independent developer testing will be required to confirm xAI’s performance claims.
Whether Grok 4.5 can displace higher-performing models such as GPT-5.6 or Opus 4.8 will depend on real-world cost, latency, and reliability during sustained agent workloads, analysts and developers told reporters in the coverage of the launch.
