Have we hit a scaling wall in base models? (non reasoning)
Grok 3 was supposedly trained on 100,000 H100 GPUs, which is in the ballpark of about 10x more than models like the GPT-4 series and Claude 3.5 Sonnet Yet they're about equal in abilities. Grok 3 isn't AGI or ASI like we hoped. In 2023 and 2024…