xAI launches Grok 4.5, positioning it as an Opus-class coding and agents model with improved efficiency

xAI publicly launched Grok 4.5 as a new frontier model focused on coding and agents, developed in partnership with Cursor. The company positioned it as an Opus-class model that is faster, more token-efficient, and lower cost than comparable offerings.

Elon Musk described Grok 4.5 internally as "roughly comparable to Opus 4.7, but much faster," emphasizing its utility for Tesla and SpaceX engineers over benchmark chasing. xAI’s official account framed Grok 4.5 as "our first model trained specifically for coding and agents," and Cursor called it "our most powerful model yet" and "the first we’ve built for more than software engineering."

Official pricing for Grok 4.5 was set at $2 per 1 million input tokens and $6 per 1 million output tokens. Artificial Analysis independently reported the same price point and noted cache hits were discounted by 75% to $0.50 per 1 million tokens. Musk also indicated the context window would likely move from 500,000 back to 1 million "by next week."

Third-party evaluations by Artificial Analysis placed Grok 4.5 at #4 on its Intelligence Index with a score of 54, behind Fable 5, GPT-5.5, and Opus 4.8. On the GDPval-AA v2 Elo, Grok 4.5 scored 1543, also ranking #4 behind Anthropic’s latest Claude releases. It achieved a top score of 33% on the τ³-Banking benchmark, surpassing GPT-5.5 (xhigh) at 31%.

Artificial Analysis’ Coding Agent Index scored Grok 4.5 at 76 in Grok Build, described as "on par with GPT-5.5 in Codex" and below Fable 5 in Claude Code. The analysis highlighted Grok 4.5’s cost efficiency, with cost per Intelligence Index task at $0.31, cost per GDPval task at $0.49, and cost per Coding Agent Index task at $2.59. Average output tokens per Intelligence Index task were approximately 14,000, over 60% lower than Opus 4.8, and average total tokens per Coding Agent Index task were 1.9 million, compared to 7.2 million for Fable 5 in Claude Code and 6.2 million for GPT-5.5 in Codex.

Cursor announced in-product availability for Grok 4.5 with "double usage for the first week." Early ecosystem support included day-zero availability in Grok Build, API, and Cursor, as well as day-zero support in Hermes Agent. Additional live availability was confirmed in Hermes Agent, Portal, OpenRouter, and Grok subscriptions.