xAI releases a new Grok model that a benchmark firm rates on par with OpenAI's flagship

xAI released Grok 4.6, an update to Grok 4.5 that the company said builds on the earlier model with a particular focus on long-running agents and more ambitious interactive and visual tasks. Pricing was unchanged from Grok 4.5, at $2 per million input tokens and $6 per million output tokens, with a faster-serving variant available at double the cost.

xAI cited a score of 61 on the Artificial Analysis Intelligence Index — a third-party benchmark aggregator, not xAI’s own measure — putting Grok 4.6 level with OpenAI’s GPT-5.6 Sol on that index. The company also reported internal or partner benchmark figures including 1753 on GDPVal-AA v2, 69.9% on CursorBench v3.2, 65.9% on DeepSWE v1.1 and 61.3% on FrontierCode v1.1; these are company-reported figures and, aside from the Artificial Analysis score, were not independently corroborated at release.

The release came alongside a rapid-fire roadmap signal from Elon Musk, who posted on X that a further model, Grok 4.7, was “significantly better than 4.6” and should be ready in three to four weeks, with initial training already complete and supplemental training under way using SpaceX company data. xAI’s pace of releases — Grok 4.5 in July, 4.6 in August, and a further model already flagged for the following month — continued a cadence of frequent, incremental updates rather than the longer gaps between major version numbers typical of rival labs.