Elon Musk just announced Grok 4.5 is in private beta at SpaceX and Tesla. New foundation model underneath it, 1.5 trillion parameters, fed with Cursor training data. The early reports say it is performing at or above Claude Opus level on hard reasoning tasks.
This is not a small update. This is a new architecture, new training, new performance tier. SpaceX is planning to drop a new foundation model every month from here on. That tells you something about the pace at which xAI is moving.
The competitive field just shifted again. Six months ago, Opus was clearly leading for hard problems. GPT-5.6 Sol came in hot but locked behind government approval. Now Grok 4.5 is matching or beating both of them, and it is available right now to the people who run SpaceX. That matters for real work, right now, today.
What is interesting is not that xAI built a competitive model. What is interesting is that they are willing to deploy it internally, test it in production at SpaceX, and iterate monthly. That is a different kind of speed than what you see at OpenAI or Anthropic. Whether it translates to public availability is still an open question. But for the people running the most complex systems on Earth, it just did.
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277
Sources:
https://www.buildfastwithai.com/blogs/ai-news-today-june-29-2026
https://x.ai/news
Repost this. Thanks.

