Reading up on llm-optimization

A speculative analysis of OpenAI's and Anthropic's fast inference strategies, claiming batch-size optimization and Cerebras hardware enable speed at the cost of model quality, though experts challenge key technical assumptions.