xAI releases Grok 4.6 with coding focus, claims parity with GPT-5.1
xAI released Grok 4.6 on Wednesday, positioning the new model around agentic coding and knowledge work. The company says Grok 4.6 matches OpenAI’s GPT-5.1 across several benchmarks, including the Artificial Analysis Intelligence Index, a composite score of nine benchmarks. The release arrives as xAI tries to rebuild credibility after earlier controversies and as it leans more visibly on its relationship with agentic coding tool Cursor.
What xAI is claiming about Grok 4.6
xAI says Grok 4.6 reaches frontier-level performance on multiple agentic coding and knowledge work benchmarks. The headline comparison is parity with GPT-5.1, particularly on the Artificial Analysis Intelligence Index, which aggregates results across nine separate evaluations rather than relying on a single test. On X, Elon Musk went further than the announcement, claiming Grok 4.6 is “objectively #1” on intelligence per speed and cost, framing that goes beyond what the benchmark match alone supports.
Why the Cursor connection matters
The release follows xAI’s deepening tie with Cursor, the agentic coding company xAI partnered with and reportedly acquired. According to the reporting, Grok 4.5 was the first xAI model trained in part on the real-world usage data Cursor had accumulated, and Grok 4.6 underwent an even longer supplemental training run on that data. The result, at least on benchmarks, is a return to the frontier tier after a stretch in which xAI appeared to lag behind OpenAI and Anthropic.
Where Grok 4.6 is available first
Grok 4.6 is rolling out in Cursor alongside Grok Build, xAI’s coding agent. xAI and Cursor also jointly released Grok Bot in beta earlier in the week, an always-on agent built through the partnership. The pairing is now a central part of how xAI is presenting the model: less a standalone chatbot and more a coding-focused stack anchored by Cursor’s distribution and usage data.
Reputation challenges that go beyond benchmarks
Performance gains do not address the public perception problems Grok carries. The model has been widely used to mass-produce non-consensual nude images, including of children, a controversy that continues to shadow the brand. Adoption inside the U.S. federal government has also been limited. Despite Musk spending roughly $400 million to help elect the second Trump administration, federal agencies have shown little appetite for Grok and several have raised safety concerns about it.
Enterprise adoption is similarly thin. According to Ramp’s AI Index, only about 4% of companies that have adopted AI tools pay for xAI’s offering, leaving Grok as a relatively small player in the business market. Cursor’s team appears to have improved Grok’s raw benchmark results, but the broader reputational work is still ahead.
FAQ
What is Grok 4.6?
Grok 4.6 is the latest publicly released model from xAI, made available on Wednesday. xAI positions it around agentic coding and knowledge work and says it matches OpenAI’s GPT-5.1 on several benchmarks, including the Artificial Analysis Intelligence Index.
How is Grok 4.6 different from Grok 4.5?
According to reporting, Grok 4.6 underwent an even longer supplemental training run on real-world usage data from Cursor than Grok 4.5 did. Grok 4.5 was the first xAI model trained in part on that Cursor dataset.
Where can people use Grok 4.6 first?
Grok 4.6 is available first in Cursor and through Grok Build, xAI’s coding agent. xAI and Cursor also released Grok Bot in beta earlier in the week as a persistent, always-on agent.
Related coverage
This article summarizes reporting from gizmodo.com.