The survey indicates a clear preference for OpenAI’s models among organizations using large language models. A majority, 58%, are using GPT-4, showcasing its dominance in the field. GPT-3.5, also from OpenAI, is used by 21%, further cementing OpenAI’s lead. Other models have a smaller share: Google AI’s PaLM 2 is at 5%, Claude v1 from Anthropic at 4%, and LaMDA at 3%. Falcon 40B and MPT-30B each hold 2%, while Cohere and Guanaco-65B from Stability AI are at 1%. Notably, newer or less known models like Vicuna 33B, Bloom, and Llama2 have minimal to no reported usage. This distribution underscores OpenAI’s significant influence in the current landscape of large language models..
