Elon Musk has provided an exciting update on his latest AI model, which is Grok 4.5. He revealed that xAI's newest model is undergoing private beta testing at both SpaceX and Tesla and mentioned early test results are promising. 

Musk announced the latest development on Twitter, stating that the Grok 4.5 was developed based on the 1.5-trillion-parameter V9 foundation model by xAI and using Cursor data. This model is now undergoing testing in a private beta at SpaceX and Tesla.

He further stated that the first test results show that Grok 4.5 works either equal to or even better than Anthropic's Cloud Opus. Moreover, the independent benchmark results were not yet published.

The v9 foundation model on which Grok 4.5 is built has 1.5 trillion parameters. This is approximately three times larger than the existing v8-small model (0.5 trillion parameters). The aim is to further improve the AI's understanding, response capabilities, and coding performance.

Musk said data from the Cursor AI coding assistant was also added to the model’s additional training. It has improved the model’s coding capabilities and ability to solve complex problems, he said.

According to Elon Musk, Grok's reinforcement learning (RL) model is constantly improving. Additionally, daily improvements are being made to the Grok build coding harness to make future models even more capable.

Musk stated that xAI plans to launch a completely freshly trained AI model every month via SpaceX for the rest of this year. The public release of the V9-Medium model was originally expected in mid-June, but private testing of Grok 4.5 is now considered the next major step.

Currently, Grok 4.5 is only available in private beta testing within SpaceX and Tesla. The company has not announced a public release date. Furthermore, independent benchmark results have not yet been released, so actual performance can only be assessed after extensive testing.