xAI Grok V9-Medium 1.5T model completes training for coding

Grok V9-Medium is xAI’s first clearly “frontier-class” move in the coding arena: a 1.5T-parameter model that triples the scale of the current production Grok V8-Small and is explicitly tuned with rich Cursor code data. That combination—large parameter count plus targeted coding corpus—signals a serious bid to compete with Anthropic’s Claude, OpenAI’s GPT-5.x line, and Google’s Gemini for developer mindshare. The announcement also clarifies months of speculation about Grok’s true scale, confirming that the widely-used model until now was only 0.5T parameters.([ai-insight.org](https://www.ai-insight.org/reports/grok-v9-medium-1500b-cursor-data-2026))

Strategically, xAI is trying to differentiate on agentic coding and integration into Musk’s broader ecosystem (X, Tesla, possibly SpaceXAI), rather than generic chat performance. Folding in Cursor’s workflow data aims to make Grok feel less like a chatbot and more like a power tool for software engineering. The promise to open‑source the 0.5T model by year-end is equally notable: it’s a play for community traction and a way to narrow the open‑source gap with Meta and Chinese labs while still keeping its best weights proprietary.([news.aibase.com](https://news.aibase.com/tw/news/28307))