Decrypt tested #Grok 4 and found political bias in the chatbot

Decrypt journalists tested Grok 4 Basic — a simplified version of the flagship Grok 4 Heavy (priced at $300/month) — to determine whether the $30 monthly price is justified.

The new base model of artificial intelligence (#AI ) Grok 4 Basic from Elon Musk, which xAI presented recently, evoked mixed impressions. The publication noted a high level of logical thinking and voice capabilities, but also a noticeable lack of creativity and political bias.

In logical and legal tasks, Grok 4 exceeded expectations. In the realm of scientific reasoning, Grok 4 operates at a PhD level, providing accurate and thorough explanations. In creative tasks, Grok 4 failed. Its stories appeared formulaic, with dull dialogues and vague construction. Despite good structural logic, narrative energy was absent. There were also problems with programming. Despite claimed improvements, Grok 4 was unable to create a fully functional game even after several iterations, and fixing one bug often led to the emergence of others.