Musk: The performance gap between Grok V8 and V9 is huge.

CoinFeed reported on May 16th that Elon Musk posted on the X platform that the newly completed Grok V9 (1.5T parameters) training run "performed very well," and this result has not yet included the supplementary training portion using Cursor data. The current internally developed base model version is V9, which has significant improvements over V8 in data cleaning, training methods, and model size, and has been optimized for the Blackwell architecture to improve computing power utilization. Musk emphasized that, in contrast, the current publicly released version v4.2 is based on the V8 base model, with a parameter size of approximately 0.5T, running on the Hopper architecture, and still has certain limitations in terms of training data quality and coverage. The performance gap between Grok V8 and V9 is huge, and the new generation model represents a leap forward in overall capabilities.