LLM Engines

A reference list of LLM and embedding models by provider: generative models, reasoners, embeddings, and rerankers. Includes specs, use cases, deployment options, and pricing.

• text-embedding-3-smallOpen AIEmbeddingText-only · API~$0.02/1M tokens

  RAG, semantic search, classification, recommendations, clustering, near-duplicate detection

• text-embedding-3-largeOpen AIEmbeddingText-only · API~$0.13/1M tokens

  RAG, semantic search, classification, recommendations, multilingual search, complex queries

• CLIPOpen AIEmbeddingText + image · LocalFree (open source)

  Text–image retrieval, image search by text, zero-shot image classification

• GPT-4oOpen AIGenerativeText + image · API~$2.50/1M input, $10/1M output

  Chat, coding, analysis, general tasks, vision

• GPT-4o miniOpen AIGenerativeText + image · API~$0.15/1M input, $0.60/1M output

  Chat, simple tasks, high volume, cost-sensitive use

• o1Open AIReasoning / thinkerText-only · API~$15/1M input, $60/1M output

  Math, coding contests, science, complex reasoning, chain-of-thought

• o1-miniOpen AIReasoning / thinkerText-only · APILower than o1

  Lightweight reasoning, cost-sensitive reasoning tasks

• BGE-large-en-v1.5Sentence Transformers (Hugging Face)EmbeddingText-only · LocalFree (open source)

  RAG, semantic search, retrieval, classification, open-source deployments

• GTE-large-en-v1.5Sentence Transformers (Hugging Face)EmbeddingText-only · LocalFree (open source)

  RAG, search, classification, multilingual

• E5-mistral-7b-instructSentence Transformers (Hugging Face)EmbeddingText-only · LocalFree (open source)

  High-quality retrieval, RAG, complex queries

• BGE-reranker-largeSentence Transformers (Hugging Face)RerankerText-only · LocalFree (open source)

  RAG reranking, passage–query relevance

• BGE-reranker-v2-m3Sentence Transformers (Hugging Face)RerankerText-only · LocalFree

  Multilingual reranking

• Gemini Embedding 2GoogleEmbeddingMultimodal · APItext: ~$0.20/1M tokens

  Multimodal RAG, cross-modal search, semantic search over text, images, video, audio, documents

• Gemini 2.0 FlashGoogleGenerativeMultimodal · Gemini API, Vertex AI~$0.10/1M input, $0.40/1M output

  Chat, fast responses, high volume, cost-sensitive use

• Gemini 2.0 ProGoogleGenerativeMultimodal · Gemini API, Vertex AIHigher than Flash

  Complex tasks, coding, analysis, higher quality

• Gemini 2.5 Pro (Thinking)GoogleReasoning / thinkerMultimodal · Gemini API, Vertex AIHigher than standard Pro

  Math, coding, multi-step reasoning, agents

• Gemini 2.5 Flash (Thinking)GoogleReasoning / thinkerMultimodal · Gemini API, Vertex AILower than Pro thinking

  Lightweight reasoning, cost-sensitive reasoning

• Vertex AI Ranking APIGoogleRerankerText-only · Vertex AIPer RAG request

  RAG reranking, semantic ranking

• embed-v4CohereEmbeddingMultimodal (text + image) · API~$0.10–0.12/1M tokens

  Multimodal RAG, text + image + PDF, mixed content

• embed-english-v3.0CohereEmbeddingText-only · API~$0.10/1M tokens

  English RAG, search, classification

• embed-multilingual-v3.0CohereEmbeddingText-only · API~$0.10/1M tokens

  Multilingual RAG, search

• Command R+CohereGenerativeText-only · API~$2.50/1M input, $10/1M output

  RAG, tool use, agents, multilingual

• Command RCohereGenerativeText-only · API~$0.15/1M input, $0.60/1M output

  RAG, chat, cost-sensitive use

• Command R7BCohereGenerativeText-only · API~$0.04/1M input, $0.15/1M output

  High volume, simple tasks

• Command A ReasoningCohereReasoning / thinkerText-only · APIHigher than Command R+

  Agents, tool use, complex reasoning

• rerank-v4.0-proCohereRerankerText-only · APIPer search

  RAG reranking, high relevance

• rerank-v3.5CohereRerankerText-only · API~$2/1,000 searches

  Multilingual reranking

• voyage-3-largeVoyage AIEmbeddingText-only · APIPremium tier

  High-accuracy RAG, semantic search, multilingual retrieval, enterprise retrieval quality

• voyage-3Voyage AIEmbeddingText-only · APIMid tier

  General-purpose RAG, semantic search, recommendations, classification

• voyage-3-liteVoyage AIEmbeddingText-only · APILow-cost tier

  High-volume embedding pipelines, cost-sensitive search/RAG

• voyage-code-3Voyage AIEmbeddingText + code · APISpecialized tier

  Code search, code RAG, repository retrieval, code similarity

• rerank-2Voyage AIRerankerText-only · APIPer-request

  Re-ranking top-k retrieved documents for higher precision in RAG

• rerank-2-liteVoyage AIRerankerText-only · APILower than rerank-2

  Cost-sensitive re-ranking at larger scale

• Qwen2.5-72B-InstructAlibaba QwenGenerativeText-only · Local or APIOpen weights or provider API

  High-quality chat, analysis, coding, complex instruction following

• Qwen2.5-32B-InstructAlibaba QwenGenerativeText-only · Local or APIOpen weights / provider API

  Strong quality with lower cost/latency than 72B

• Qwen2.5-14B-InstructAlibaba QwenGenerativeText-only · Local or APIOpen weights / provider API

  Mid-size production assistants, cost-aware coding/chat

• Qwen2.5-7B-InstructAlibaba QwenGenerativeText-only · Local or APIOpen weights / provider API

  Lightweight assistants, edge/server cost-sensitive workloads

• Qwen2.5-Coder-32B-InstructAlibaba QwenGenerative (code-focused)Text-only · Local or APIOpen weights / provider API

  Code generation, refactoring, repo Q&A, code reasoning

• Qwen2.5-VL-72B-InstructAlibaba QwenMultimodal generativeText + image · Local or APIOpen weights / provider API

  Vision + text understanding, document/image question answering

• DeepSeek-V3DeepSeekGenerativeText-only · APIProvider-dependent

  General chat, coding, analysis, high-quality assistant tasks

• DeepSeek-R1DeepSeekReasoning / thinkerText-only · APIProvider-dependent

  Reasoning-heavy tasks, math, logic, multi-step planning

• DeepSeek-R1-Distill-Llama-70BDeepSeekReasoning / thinkerText-only · Local or APIOpen weights or provider API

  Cost-aware reasoning, self-hosted reasoning workloads

• DeepSeek-R1-Distill-Qwen-32BDeepSeekReasoning / thinkerText-only · Local or APIOpen weights or provider API

  Mid-size reasoning deployments, balanced quality/latency

• DeepSeek-Coder-V2DeepSeekGenerative (code-focused)Text-only · Local or APIOpen weights or provider API

  Code generation, refactoring, debugging, repo-level coding assistance

• DeepSeek-EmbeddingDeepSeekEmbeddingsText-only · APIProvider-dependent

  RAG, semantic search, retrieval indexing, clustering

• Kimi (flagship)Moonshot AI (Kimi)GenerativeText-only · APIProvider-dependent

  General chat, analysis, long-document QA, productivity assistant

• Kimi Long-ContextMoonshot AI (Kimi)Generative (long-context)Text-only · APIProvider-dependent

  Very large documents, long conversations, report synthesis

• Kimi Lightweight / FastMoonshot AI (Kimi)GenerativeText-only · APILower-cost tier

  High-volume chat, latency-sensitive assistants, cost-aware production

• Kimi Reasoning-OrientedMoonshot AI (Kimi)Reasoning / thinkerText-only · APIProvider-dependent

  Multi-step reasoning, math/logic-heavy prompts, complex planning

• Kimi Vision-CapableMoonshot AI (Kimi)Multimodal generativeText + image · APIProvider-dependent

  Image + text understanding, document screenshots, charts/slides QA