OpenAI has released a new ChatGPT endpoint with significant improvements in speed and cost. This new API presents a different interface compared to the existing GPT-3 endpoint, prompting LangChain to adapt its abstractions to accommodate not only ChatGPT but also future chat-based models. The new ChatGPT API allows for differentiation between "user", "assistant", and "system" messages, which could improve the safety of applications built on top of these models. LangChain's new abstractions include those for different types of chat messages, chat models, chat message prompt templates, and prompt values to ensure seamless interoperability between language model providers. The ultimate goal is to allow users to fully take advantage of the new chat model interface while maintaining compatibility with existing LLM APIs.