When we first deployed LibreChat for enterprise clients, OpenAI (Azure OpenAI) was the first model provider on the platform. That was by design — the ChatGPT models were the obvious starting point, and they worked well enough.
Last fall, we added Anthropic's Claude models. We also added Google Gemini. Both were available alongside GPT, and we were curious to see how clients would respond to having a real choice.
The migration to Claude started almost immediately. Users exploring model options gravitated to Claude (particularly Claude Opus) and stayed there. Agents we'd built on GPT continued to run as designed, but for direct model usage, the preference was clear and consistent. At users' request, we moved the coding agent we had built from GPT-5.2 to Claude Opus 4.5, and usage skyrocketed.
By early March, Claude accounted for over 52% of model usage at our largest client, and that share continues to grow at the expense of GPT usage. Interestingly, Gemini has not been a significant factor.
We didn't engineer this outcome. We gave clients a choice and watched what happened. The data made the decision for us: Claude is now our platform default.
We have a lot more planned for Claude in the coming months - stay tuned!