xAI officially announces the release of the latest version of Grok 4.6, which builds on Grok 4.5 with a particular focus on long-running agents and more challenging interactive and visual tasks. Grok 4.6 can handle complex tasks across multiple steps, whether researching topics, analyzing information, working across codebases, or turning ideas into polished applications or work products.
Grok 4.6 Achieves Frontier Intelligence Across Multiple Benchmarks
Grok 4.6 achieves frontier intelligence across multiple agent coding and knowledge work benchmarks, and matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, a composite score across nine benchmarks.
"Grok 4.6 achieves frontier intelligence across multiple agent coding and knowledge work benchmarks."
xAI
Grok 4.6's Training Process
Grok 4.6 underwent longer supplementary training than Grok 4.5, using curated model-generated data to enhance reasoning and advanced technical concept capabilities, and improved the optimizer and training recipe. This provided a stronger foundation for subsequent SFT and RL stages.
Turning Ambitious Ideas into Actual Projects
During testing of Grok 4.6, the model proved particularly strong at turning broad product ideas into working initial versions. It can research unfamiliar domains, structure applications, implement core interactions, and continuously refine results through multi-turn feedback.
"Grok 4.6 produces stronger initial results on visual and interactive projects than Grok 4.5."
xAI
Improvements in Safety and Capabilities
Grok 4.6's safety measures have been improved and calibrated according to the model's capabilities. The safety stack aims to maximize utility and safety for legitimate use cases, enabling Grok 4.6 to function in areas such as vulnerability patching, accelerating engineering design cycles, and enhancing AI research.
"Our safety evaluation work reflects Grok 4.6's expanded capabilities."
xAI
Grok 4.6 is now available in Cursor and Grok Build, and via API and other partners including OpenRouter, Vercel, and Cloudflare. Pricing starts at $2 per million input tokens (approx. 15.6 HKD) and $6 per million output tokens (approx. 46.8 HKD), with a double-usage offer in Grok Build and Cursor for the first week.
Source: xAI Official Announcement