YBX AI Radar
AI Knowledge YBX Data Page

Grok V9 Skipped GitHub and Trained on How You Actually Code in Cursor, All 1.5 Trillion Parameters

Author: ybx-ai-radar
AI Radar Summary

This article from Towards AI details Grok V9, a 1.5-trillion-parameter code-focused AI model developed by xAI. Unlike traditional code AI models that train on scraped public GitHub code, Grok V9 was trained on real in-editor coding behavior data from the Cursor editor, aligning more closely with actual developer workflows.

Source Towards AI
Original Time Jun 16, 2026 15:58 GMT+8
Importance Score 8.0 / 10
Related Entities xAI, Grok V9, Cursor, GitHub, Towards AI
Grok V9 Skipped GitHub and Trained on How You Actually Code in Cursor, All 1.5 Trillion Parameters

One-Sentence Explanation

Grok V9 is a 1.5-trillion-parameter code-focused large language model developed by xAI, which abandoned the conventional practice of scraping public GitHub code for training and instead used real coding behavior data from the Cursor editor.

Plain Language Understanding

Most existing code AI models learn by copying public code snippets from GitHub, similar to copying homework from others. Grok V9, however, learns from the actual coding processes of developers using the Cursor editor, such as how they debug, complete code, and handle real development issues, making its training more aligned with real-world development scenarios.

Application Scenarios

  • Code completion and bug fixing for professional developers
  • Generation of complex business code
  • Code learning assistance for novice developers

xAI, Grok series models, Cursor editor, large code models, model parameter scale

YBX AI Radar

Related Reading