Over the past few days we did an investigation of the main LLM providers, and have observed up to a 40% difference in average speed (tokens / second) from the leading LLM providers like GPT4.