xAI released Grok 4.6 publicly on Wednesday, positioning it as a frontier model for agentic coding and knowledge work. The release marks a return to the top tier of AI benchmarks for xAI, which had lagged behind competitors like OpenAI and Anthropic, and comes shortly after the company partnered with the agentic coding startup Cursor.<\/p>\n
xAI says Grok 4.6 delivers frontier intelligence across several agentic coding and knowledge work benchmarks. The company also reports that the model matches OpenAI’s GPT-5.6 Sol on a range of evaluations, including the Artificial Analysis Intelligence Index, a composite score drawn from nine benchmarks.<\/p>\n
On X, Elon Musk went further, claiming that Grok 4.6 is “objectively #1 when considering intelligence, speed cost.” That framing is notable given that the launch announcement’s strongest claim is parity with a competitor rather than a clear win.<\/p>\n
Grok’s repositioning follows xAI’s partnership with and possible acquisition of Cursor, an agentic coding company. Grok 4.5 was the first version of xAI’s model trained in part on real-world usage data Cursor had accumulated, and Grok 4.6 reportedly underwent an even longer supplemental training run using that same data.<\/p>\n
The result, at least on xAI’s benchmarks, is a model that competes again rather than one that risks fading from the frontier conversation. xAI is leaning into the relationship: Grok 4.6 is available first in Cursor, alongside Grok Build, the xAI coding agent, starting on release day. Cursor and xAI also jointly released an always-on agent called Grok Bot in beta earlier in the week.<\/p>\n
Grok’s prior reputation is heavily shaped by misuse. Earlier versions were used to mass-produce non-consensual nude images, including of children, a controversy that xAI has struggled to move past. Multiple federal agencies have raised concerns about the model’s safety, and the second Trump administration, which Musk spent $400 million to support, has shown limited interest in adopting Grok.<\/p>\n
Enterprise adoption is also thin. According to Ramp’s AI Index, xAI’s option has been selected by just 4% of companies that have adopted AI tools, leaving Grok as a niche player in the enterprise market.<\/p>\n
For benchmark watchers, Grok 4.6 is a credible re-entry point. xAI now has a coding-focused distribution channel in Cursor, real-world usage data feeding its training pipeline, and benchmark numbers that, on paper at least, put it in the same conversation as OpenAI. Whether that performance translates into broader adoption depends less on scores and more on the trust issues that have defined Grok’s public life so far.<\/p>\n
Grok 4.6 is the latest publicly released AI model from xAI, made available on Wednesday. xAI says it achieves frontier intelligence on several agentic coding and knowledge work benchmarks and matches OpenAI’s GPT-5.6 Sol on evaluations including the Artificial Analysis Intelligence Index, a composite score of nine benchmarks.<\/p>\n
xAI partnered with and may have acquired Cursor, an agentic coding company whose real-world usage data has been folded into Grok’s training. Grok 4.5 was the first xAI model trained in part on Cursor’s data, and Grok 4.6 reportedly underwent a longer supplemental training run using the same data. Grok 4.6 and Grok Build are available first inside Cursor, and the two companies jointly released Grok Bot in beta.<\/p>\n
Grok has been associated with the mass generation of non-consensual nude images, including of children, and multiple federal agencies have raised safety concerns. According to Ramp’s AI Index, only 4% of companies that have adopted AI tools pay for xAI’s option, leaving Grok with limited enterprise traction despite its new benchmark performance.<\/p>\n