Interesting that Grok 4.5 nears or exceeds OpenAI and Anthropic models in some benchmarks at only 1.5 trillion parameters per their announcement post.