Grok V9 nears release after training

This is a meaningful scale jump if the numbers hold, but the bigger story is data quality and productization, not just parameter count. xAI is signaling that Grok’s next step is meant to matter for coding and agent workflows, where model usefulness shows up fast and hype gets punished quickly.

• –A 1.5T run is a serious compute bet, suggesting xAI is still competing on raw model scale as much as product features
• –The Cursor-data angle is more interesting than the size claim, because code-specific data is what could move Grok closer to the coding-agent tier
• –Musk’s wording implies the model is trained, but the supplemental Cursor data is not yet in this run, so the integration story is still forward-looking
• –The current public Grok line already spans Grok 4.2, connectors, voice, imagine, and API products, so V9 looks aimed at improving the core model under all of those surfaces
• –Because the update comes from a post rather than a formal xAI launch note, the claim is strong on signal but still weak on independently verified product detail