Without changing its foundation, SpaceXAI pushes Grok to a level close to GPT-5.6 Sol thanks to additional training targeted on agentic tasks.
SpaceXAI, born from the merger of xAI and SpaceX, launched Grok 4.6 on August 12, an update that retains the foundation of Grok 4.5 but adds extended additional training, regenerated fine-tuning trajectories, and reinforcement learning in agentic environments. The model manages a 500,000-token context, accepts text and image input, and introduces a new level of "xhigh" reasoning in addition to the low/medium/high levels.
On Artificial Analysis' Intelligence Index, Grok 4.6 joins GPT-5.6 Sol and is hot on the heels of Claude Fable 5, with a five-point gain over Grok 4.5 in just over a month. It is available immediately in Grok Build, Cursor, Grok Bot, and via the API.
Read the full article on the original source
Go to the source →