What is a framework page?
It says where that library stores the model id, how to change it, and which upcoming OpenAI chat shutdowns are in this dataset. Framework facts link the official docs. Dates come from the model records.
Frameworks
These pages say where a library stores the model id: a constructor argument, a config field, a node field, or a proxy alias. Swap that string when the id is in the shutdown table. The table is built from SunsetDiff records. Framework facts link the official docs. Last reviewed . Publisher: SunsetDiff.
LiteLLM and OpenRouter fallback samples stay on How to swap a deprecated model. GPT-6 function tools that Chat Completions rejects are on Move function calling to the Responses API and the request errors.
| Framework | Where the model id lives |
|---|---|
| LangChain | The OpenAI model id is the model argument on ChatOpenAI. On Azure that string is the deployment name. |
| LiteLLM | completion() takes a model string. The proxy alias is model_name. The provider id is litellm_params.model. |
| Vercel AI SDK | The model id is the string passed to openai(), openai.chat(), or openai.responses(). |
| LlamaIndex | The model id is the model argument of llama_index.llms.openai.OpenAI. Settings.llm is the global default. |
| OpenRouter | The request field model is the slug. A ~ prefix is a latest alias. models is the fallback list. |
| Pipecat | OpenAILLMService is Chat Completions. The settings default model is gpt-4.1. Responses is a different class. |
| n8n | The Model field on the OpenAI Chat Model node is the model. Chat Completions is the default until you turn on Use Responses API. |
| Langflow | The core component uses provider and model_name. The OpenAI bundle component uses model. |
Questions
Does the swap guide still exist?
Yes. How to swap a deprecated model keeps the LiteLLM and OpenRouter fallback samples. These pages add Pipecat, n8n, LlamaIndex, and Langflow, and the shutdown table.
Who publishes these pages?
Publisher: SunsetDiff. The site is independent and not affiliated with any AI provider or framework.