Title: Modelle und Reranker f%C3%BCr RAG %E2%80%94 LLM EU Canonical: https://llmeu.com/de/use/rag/ Markdown mirror: https://llmeu.com/de/use/rag.md Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json # Modelle und Reranker f%C3%BCr RAG Modelle und Reranker f%C3%BCr retrieval-augmentierte Generierung. Bei RAG entscheiden Retriever und Reranker oft die Antwort; deshalb stehen Embedding- und Reranking-Modelle mit in dieser Liste. 65 models %C2%B7 [Im Katalog filtern](/de/models/?task=rag) - [Codestral 25.08](/de/models/codestral-25-08/) %E2%80%94 Mistral AI %C2%B7 128K %C2%B7 proprietary - [Command A+ (05-2026)](/de/models/command-a-plus-05-2026/) %E2%80%94 Cohere %C2%B7 218.75B %C2%B7 128K %C2%B7 apache-2.0 - [Command R (08-2024)](/de/models/command-r-08-2024/) %E2%80%94 Cohere %C2%B7 32.3B %C2%B7 128K %C2%B7 cc-by-nc-4.0 - [DeepSeek OCR](/de/models/deepseek-ocr/) %E2%80%94 DeepSeek %C2%B7 3.3B %C2%B7 mit - [DeepSeek OCR 2](/de/models/deepseek-ocr-2/) %E2%80%94 DeepSeek %C2%B7 3.4B %C2%B7 apache-2.0 - [DeepSeek-V3.2](/de/models/deepseek-v3-2/) %E2%80%94 DeepSeek %C2%B7 685.36B %C2%B7 160K %C2%B7 mit - [DeepSeek-V4-Flash (0731)](/de/models/deepseek-v4-flash-0731/) %E2%80%94 DeepSeek %C2%B7 304.18B %C2%B7 1024K %C2%B7 mit - [Devstral Small 2 (24B)](/de/models/devstral-small-2-24b/) %E2%80%94 Mistral AI %C2%B7 24.01B %C2%B7 256K %C2%B7 apache-2.0 - [EmbeddingGemma 300M](/de/models/embeddinggemma-300m/) %E2%80%94 Google DeepMind %C2%B7 0.3B %C2%B7 2K %C2%B7 gemma - [EuroLLM-9B-Instruct (2512)](/de/models/eurollm-9b-instruct-2512/) %E2%80%94 EuroLLM (utter-project) %C2%B7 9.15B %C2%B7 32K %C2%B7 apache-2.0 - [GLM-4 9B Chat 1M](/de/models/glm-4-9b-chat-1m/) %E2%80%94 Z.ai (Zhipu AI) %C2%B7 9.5B %C2%B7 other - [GLM-5.3](/de/models/glm-5-3/) %E2%80%94 Z.ai (Zhipu AI) %C2%B7 753.33B %C2%B7 1024K %C2%B7 other - [GLM-OCR](/de/models/glm-ocr/) %E2%80%94 Z.ai (Zhipu AI) %C2%B7 1.33B %C2%B7 128K %C2%B7 mit - [gpt-oss-120b](/de/models/gpt-oss-120b/) %E2%80%94 OpenAI %C2%B7 116.83B %C2%B7 128K %C2%B7 apache-2.0 - [gpt-oss-20b](/de/models/gpt-oss-20b/) %E2%80%94 OpenAI %C2%B7 20.91B %C2%B7 128K %C2%B7 apache-2.0 - [Granite 3.1 8B Instruct](/de/models/granite-3-1-8b-instruct/) %E2%80%94 IBM %C2%B7 8.2B %C2%B7 apache-2.0 - [Granite 3.2 8B Instruct](/de/models/granite-3-2-8b-instruct/) %E2%80%94 IBM %C2%B7 8.2B %C2%B7 apache-2.0 - [Granite 3.3 8B Instruct](/de/models/granite-3-3-8b-instruct/) %E2%80%94 IBM %C2%B7 8.2B %C2%B7 apache-2.0 - [Granite 4.0 3B Vision](/de/models/granite-4-0-3b-vision/) %E2%80%94 IBM %C2%B7 4B %C2%B7 apache-2.0 - [Granite 4.1 30B](/de/models/granite-4-1-30b/) %E2%80%94 IBM %C2%B7 28.9B %C2%B7 apache-2.0 - [Granite 4.1 8B](/de/models/granite-4-1-8b/) %E2%80%94 IBM %C2%B7 8.8B %C2%B7 apache-2.0 - [Granite 4.2 30B](/de/models/granite-4-2-30b/) %E2%80%94 IBM %C2%B7 29.3B %C2%B7 apache-2.0 - [Granite 4.2 8B](/de/models/granite-4-2-8b/) %E2%80%94 IBM %C2%B7 8.8B %C2%B7 apache-2.0 - [Granite Docling 258M](/de/models/granite-docling-258m/) %E2%80%94 IBM %C2%B7 0.3B %C2%B7 apache-2.0 - [Granite Vision 4.1 4B](/de/models/granite-vision-4-1-4b/) %E2%80%94 IBM %C2%B7 4B %C2%B7 apache-2.0 - [gte-multilingual-reranker-base](/de/models/gte-multilingual-reranker-base/) %E2%80%94 Alibaba Qwen %C2%B7 0.31B %C2%B7 8K %C2%B7 apache-2.0 - [HunyuanOCR](/de/models/hunyuanocr/) %E2%80%94 Tencent %C2%B7 1.1B %C2%B7 other - [Llama 3.1 8B Instruct](/de/models/llama-3-1-8b-instruct/) %E2%80%94 Meta %C2%B7 8.03B %C2%B7 128K %C2%B7 llama3.1 - [Llama 3.3 70B Instruct](/de/models/llama-3-3-70b-instruct/) %E2%80%94 Meta %C2%B7 70.55B %C2%B7 128K %C2%B7 llama3.3 - [Llama 4 Scout 17B-16E Instruct](/de/models/llama-4-scout-17b-16e-instruct/) %E2%80%94 Meta %C2%B7 108.64B %C2%B7 10240K %C2%B7 other - [Llama Embed Nemotron 8B](/de/models/llama-embed-nemotron-8b/) %E2%80%94 NVIDIA %C2%B7 other - [Llama Nemotron Embed 1B V2](/de/models/llama-nemotron-embed-1b-v2/) %E2%80%94 NVIDIA %C2%B7 other - [Llama Nemotron Rerank 1B v2](/de/models/llama-nemotron-rerank-1b-v2/) %E2%80%94 NVIDIA %C2%B7 1.24B %C2%B7 8K %C2%B7 other - [Llama Nemotron Rerank Vl 1B V2](/de/models/llama-nemotron-rerank-vl-1b-v2/) %E2%80%94 NVIDIA %C2%B7 other - [Mistral NeMo 12B Instruct](/de/models/mistral-nemo-12b/) %E2%80%94 Mistral AI %C2%B7 12.25B %C2%B7 128K %C2%B7 apache-2.0 - [multilingual-e5-large](/de/models/multilingual-e5-large/) %E2%80%94 Microsoft %C2%B7 0.56B %C2%B7 1K %C2%B7 mit - [NVIDIA Nemotron 3 Embed 1B](/de/models/nemotron-3-embed-1b/) %E2%80%94 NVIDIA %C2%B7 other - [NVIDIA Nemotron 3 Embed 8B](/de/models/nemotron-3-embed-8b/) %E2%80%94 NVIDIA %C2%B7 other - [NVIDIA Nemotron 3 Nano 4B](/de/models/nemotron-3-nano-4b/) %E2%80%94 NVIDIA %C2%B7 3.97B %C2%B7 256K %C2%B7 other - [NVIDIA Nemotron 3 Super 120B-A12B](/de/models/nemotron-3-super-120b-a12b/) %E2%80%94 NVIDIA %C2%B7 123.61B %C2%B7 256K %C2%B7 other - [NVIDIA NV-Embed-v2](/de/models/nv-embed-v2/) %E2%80%94 NVIDIA %C2%B7 cc-by-nc-4.0 - [OlmOCR 2 7B 1025](/de/models/olmocr-2-7b-1025/) %E2%80%94 Allen Institute for AI (Ai2) %C2%B7 8.3B %C2%B7 apache-2.0 - [OlmOCR 7B 0225 Preview](/de/models/olmocr-7b-0225-preview/) %E2%80%94 Allen Institute for AI (Ai2) %C2%B7 8.3B %C2%B7 apache-2.0 - [Pharia-1-Embedding-4608-control](/de/models/pharia-1-embedding-4608-control/) %E2%80%94 Aleph Alpha %C2%B7 7.03B %C2%B7 other - [Pharia-1-LLM-7B-control-aligned](/de/models/pharia-1-llm-7b-control-aligned/) %E2%80%94 Aleph Alpha %C2%B7 7.04B %C2%B7 8K %C2%B7 other - [Phi 3 Mini 128K Instruct](/de/models/phi-3-mini-128k-instruct/) %E2%80%94 Microsoft %C2%B7 3.8B %C2%B7 mit - [Phi 3 Vision 128K Instruct](/de/models/phi-3-vision-128k-instruct/) %E2%80%94 Microsoft %C2%B7 4.1B %C2%B7 mit - [Phi-4](/de/models/phi-4/) %E2%80%94 Microsoft %C2%B7 14.66B %C2%B7 16K %C2%B7 mit - [Pixtral 12B](/de/models/pixtral-12b/) %E2%80%94 Mistral AI %C2%B7 12B %C2%B7 128K %C2%B7 apache-2.0 - [Qianfan OCR](/de/models/qianfan-ocr/) %E2%80%94 Baidu %C2%B7 4.7B %C2%B7 apache-2.0 - [Qwen3 Embedding 0.6B](/de/models/qwen3-embedding-0-6b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 Embedding 4B](/de/models/qwen3-embedding-4b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 Embedding 8B](/de/models/qwen3-embedding-8b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 Reranker 0.6B](/de/models/qwen3-reranker-0-6b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 Reranker 4B](/de/models/qwen3-reranker-4b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 Reranker 8B](/de/models/qwen3-reranker-8b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 VL Embedding 2B](/de/models/qwen3-vl-embedding-2b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 VL Embedding 8B](/de/models/qwen3-vl-embedding-8b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 VL Reranker 2B](/de/models/qwen3-vl-reranker-2b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3 VL Reranker 8B](/de/models/qwen3-vl-reranker-8b/) %E2%80%94 Alibaba Qwen %C2%B7 apache-2.0 - [Qwen3-Coder-30B-A3B-Instruct](/de/models/qwen3-coder-30b-a3b-instruct/) %E2%80%94 Alibaba Qwen %C2%B7 30.53B %C2%B7 256K %C2%B7 apache-2.0 - [Qwen3-VL-8B-Instruct](/de/models/qwen3-vl-8b-instruct/) %E2%80%94 Alibaba Qwen %C2%B7 8.77B %C2%B7 256K %C2%B7 apache-2.0 - [Unlimited OCR](/de/models/unlimited-ocr/) %E2%80%94 Baidu %C2%B7 3.3B %C2%B7 mit - [WeVisDoc 2B](/de/models/wevisdoc-2b/) %E2%80%94 Tencent %C2%B7 2.4B %C2%B7 apache-2.0 - [Yi 9B 200K](/de/models/yi-9b-200k/) %E2%80%94 01.AI %C2%B7 8.8B %C2%B7 apache-2.0 [open-weights](/de/use/open-weights/) %C2%B7[eu-builders](/de/use/eu-builders/) %C2%B7[german](/de/use/german/) %C2%B7[french](/de/use/french/) %C2%B7[long-context](/de/use/long-context/) %C2%B7[coding](/de/use/coding/) %C2%B7[vision](/de/use/vision/) %C2%B7[audio](/de/use/audio/) %C2%B7[apache-2](/de/use/apache-2/) %C2%B7[self-hostable](/de/use/self-hostable/) %C2%B7[catalogue-only](/de/use/catalogue-only/)