{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreieyn642rhspmm3evx2ewckaari7rowcbsks42w6ybaqb7zxcvuq7i",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mqmj6hmtyh32"
  },
  "path": "/t/which-embedding-model-works-best-for-rag-when-using-gpt-5-4-mini-nano-or-gpt-4-1-mini-nano-as-the-llm/1386850#post_1",
  "publishedAt": "2026-07-14T14:20:07.000Z",
  "site": "https://community.openai.com",
  "textContent": "Hi everyone,\n\nI’m building a Retrieval-Augmented Generation (RAG) system and I’m trying to choose the right embedding model.\n\nMy generation models will likely be one of these:\n\n  * GPT-5.4 mini / nano\n\n  * GPT-4.1 mini / nano\n\n\n\n\nI understand that the embedding model and the LLM are separate components in a RAG pipeline, but I would like to know what embedding models provide the best retrieval quality when paired with these models.\n\nSome questions:\n\n  1. **Embedding model selection**\n\n     * Which embedding models are currently recommended for high-quality RAG?\n\n     * Is using OpenAI embedding models the best choice when the generation model is also OpenAI?\n\n     * Are there open-source embedding models that can outperform hosted embedding models for specific use cases?\n\n  2. **Embedding dimension and performance**\n\n     * How much does embedding dimension affect retrieval quality?\n\n     * Is a larger embedding model always better, or can smaller embeddings provide similar results with lower cost and latency?\n\n  3. **Compatibility with GPT models**\n\n     * Does the choice of embedding model significantly impact the performance of GPT-5.4 mini/nano or GPT-4.1 mini/nano?\n\n     * Are there recommended combinations of embedding model + GPT model for RAG applications?\n\n  4. **Practical RAG considerations**\n\n     * For a production RAG system with many documents and frequent queries, what should be prioritized:\n\n       * Better semantic retrieval quality?\n\n       * Lower embedding cost?\n\n       * Faster retrieval latency?\n\n       * Smaller vector storage size?\n\n\n\n\nExample use case:\n\n  * Multi-tenant SaaS knowledge base\n\n  * Documents: FAQs, product information, support documents, policies\n\n  * Vector database: Pinecone / Chroma / FAISS\n\n  * LLM: GPT-5.4 mini/nano or GPT-4.1 mini/nano\n\n\n\n\nI would appreciate recommendations based on real-world RAG deployments and benchmarks.\n\nThanks!",
  "title": "Which embedding model works best for RAG when using GPT-5.4 mini/nano or GPT-4.1 mini/nano as the LLM?"
}