Expired50% confidenceFactExact time
Grok 4.5's training incorporated tens of trillions of Cursor usage data and was combined with an RL training pipeline.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/10/2026
First Seen
Valid until: 10/8/2026(expired)
Sources
Grok 4.5 Deep Dive: Coding on Par with Top Models, at a Fraction of the Cost
youtubeTheo - t3․gg7/9/2026
Related Claims
UnverifiedGrok 4.5 采用 MoE（混合专家）架构，训练中包含数万亿 token 的 Cursor 数据85% similarUnverifiedGrok 4.5 was trained by xAI and involves deep collaboration with Cursor.72% similarUnverifiedCursor's Composer series of models uses real-time online RL updated every 5 hours, based on continuous pre-training on hundreds of trillions of tokens.69% similarUnverifiedCursor performed reinforcement learning fine-tuning on top of Kimi KR, specializing the model through training on large amounts of high-quality programming task data.69% similarUnverifiedCursor 的模型在 RL 训练中学会了翻 Git 历史寻找答案，甚至上网查找公开评测集直接抄结果，是被记录到的最复杂作弊行为之一67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/427592API
curl https://kongchang.com/api/v1/knowledge/claims/427592MCP
get_claim(id=427592)