xAI launches Grok 4.5, touts cheaper AI for coding and work

xAI on Wednesday introduced Grok 4.5, pitching the model as its smartest yet and a cheaper option for coding, agentic tasks and routine knowledge work. Elon Musk said it was an “Opus-class model, but faster, more token-efficient and lower cost,” putting the company’s latest release directly into the fight for enterprise AI spending.

The pricing is the sharpest part of the sales pitch. xAI said Grok 4.5 costs $2 per million input tokens and $6 per million output tokens, while also claiming “twice greater token efficiency” than other leading models. Those numbers matter because token usage drives cost for developers and businesses that run large volumes of prompts, code generation and back-office automation. xAI framed the model as cheaper than Anthropic’s Opus pricing, signaling that Musk wants Grok judged not just by benchmark performance but by the economics of everyday deployment.

Musk said Grok 4.5 would be made public on Thursday, July 9, 2026, after what he called strong positive feedback from beta customers. The release came after xAI went public several weeks earlier, making Grok 4.5 the company’s first major model launch since the IPO. That timing gives the product extra weight inside xAI’s broader push to persuade businesses that its models can move beyond novelty chats and into paid work.

The company is leaning hard into that message. In recent months, xAI has expanded Grok into Microsoft Word and PowerPoint, Databricks, Amazon Bedrock and coding tools including Grok Build and Cursor. Grok 4.5 follows Grok 4, which launched on July 9, 2025, as xAI’s previous step-up in reasoning and tool use. The new model suggests the company is now competing on two fronts at once: raw capability and the cost structure that determines whether companies can afford to use frontier AI at scale.