We interviewed 30+ enterprise customers of LLM developers, with annual spend up into the millions, and the vast majority are leveraging models from multiple developers for their use cases.
Those include models from private players like OpenAI and Anthropic as well as Google’s Gemini and Meta’s Llama.
"Right now, we are serving 16-plus different generative AI scenarios for which we are using a mix of different LLMs. For about 50% of use cases, we use OpenAI, for 30%-40% of use cases, we use our internal fine-tuned and pre-trained LLMs based on open-source LLMs like Llama 2 or Mistral, and for others we use Anthropic or Cohere."— Senior Director of Engineering at Fortune 500 company
Enterprises are choosing models that best fit their use case, optimizing for performance and cost.
Models as specialized “coworkers”
Balancing cost and versatility
Infrastructure costs matter
Other reasons companies are working with multiple LLM developers include diversification, to reduce dependency on a single developer, and future-proofing by exposing themselves to new innovation as new models are introduced.
