Grok models performance raises questions about their effectiveness compared to Anthropic and OpenAI.