I’m not quite sure whether SuperGrok and Grok Heavy have different capabilities. Even though they both use the Grok 4.6 xHigh model, SuperGrok often gets stuck in a loop of repetitive responses and completely fails to understand the conversation. Heavy, however, works normally. I’m not using a relay or proxy—I’m logging in directly with the account. Also, SuperGrok doesn’t like executing tasks directly, while Grok Heavy carries them out and verifies the results on its own.
The gap in SuperGrok’s capabilities is way too large. It often doesn’t carry out the task or think things through, and just continues from its previous answer. What’s the reason for this?
Grok gets dumber way too quickly. Every time a new model works fine for a few days, it suddenly becomes much less intelligent, which really hurts its reputation.