ModelRefs / Retrieval-Augmented Generation (RAG) — Canonical Workflow
Retrieval-Augmented Generation (RAG) — Canonical Workflow
Canonical RAG workflow: compatible models, vector databases, frameworks, benchmarks and deployment patterns.
Overview
Retrieval-Augmented Generation (RAG) grounds a model's answers in your own documents or knowledge base by retrieving relevant context from a vector database before generation — the production-default pattern for trustworthy Q&A, internal search, and customer-facing assistants where unsupported answers are unacceptable.
Use this page to check when RAG fits your use case, which models, vector databases, and frameworks are compatible, and which architecture and benchmark references to review before committing to a retrieval stack.
Workflow fit here is provisional decision support based on goal, data, constraints, risk, and available evidence — it does not guarantee retrieval quality or answer accuracy. Evaluate the candidate stack on representative queries, including out-of-scope questions and stale or missing documents, before production use.
Implementation profile
| Category | llms |
|---|---|
| Implementation maturity | production |
| Evidence status | partial |
| Primary use cases | rag, embeddings |
| Deployment options | managed-api, self-hosted, hybrid |
| Architectures | serverless-api, managed-container, hybrid-private-cloud |
Candidate models with published references
- BGE-M3
- GPT-5
- GPT-5 Mini
- Claude Opus 4
- Llama 4 Scout
- DeepSeek R1
- Mistral Large 2
- Command R+
- o3
- o4 Mini
- Text Embedding 3 Large
- Claude Sonnet 4
Coverage means the model is a candidate worth evaluating for this workflow, not a ranking or a recommendation. Models whose reference pages are still in review are omitted.
Benchmarks relevant to this workflow
miracl, mkqa, mldr, swe-bench, aider-polyglot, gpqa, aime-2025, tau-bench, browsecomp-long-context, longfact-concepts, terminal-bench, mmmu, mmlu-pro, livecodebench.
Relevance is a coverage signal from the canonical registry. Each benchmark only describes its own protocol and date, so confirm the harness matches your workload before treating a score as evidence.
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Retrieval-Augmented Generation (RAG) — Canonical Workflow.