# NVIDIA NIM APIs NVIDIA NIM APIs is a platform for exploring and using model endpoints to build enterprise generative AI applications. It also provides blueprints and device-specific playbooks for moving from model selection to application workflows. - Findings24: - Site: - Preços: Freemium - Categorias: Plataformas de inferência de IA, Implantação de modelos de IA, Hubs de modelos de IA ## O que é NVIDIA NIM APIs? NVIDIA NIM APIs is NVIDIA’s model and application-building hub for enterprise generative AI development. It combines a searchable model catalog with blueprints and device-specific playbooks so developers can evaluate models, access available endpoints, and build AI applications from documented starting points. The model catalog spans language, multimodal, vision, speech, image, video, coding, retrieval, and specialized workloads. Models can be filtered by provider, publisher, use case, availability, and hardware-related attributes. The site also presents NVIDIA-optimized models, partner endpoints, launchable options, and models with download availability. Blueprints provide broader application patterns with workflows and code samples, while playbooks provide step-by-step guides optimized for selected machines and use cases. Available playbook categories include serving models, running LLM inference, fine-tuning models, running domain-specific workloads, building AI applications, and running AI agents. ## O que NVIDIA NIM APIs pode fazer? ### Model catalog with practical filters Browse models by provider, publisher, use case, endpoint or download availability, and related hardware attributes. The catalog includes workloads such as OCR, speech-to-text, translation, image generation, semantic search, coding, and agentic reasoning. ### NVIDIA and partner model options The catalog includes NVIDIA-published models alongside models from other publishers and inference providers. Availability indicators distinguish options such as NVIDIA optimization, free endpoints, partner endpoints, and downloadable models. ### Blueprint-based application starting points Blueprints provide workflows and code samples for building applications from the ground up. Examples cover enterprise search, AI agents, voice applications, financial services, healthcare, retail, media, robotics, and infrastructure operations. ### Device-oriented playbooks Playbooks are organized for DGX Spark, DGX Station, and RTX Workstation environments. Their stated topics include serving models, LLM inference, fine-tuning, domain-specific workloads, AI agents, and application development. ### Support for varied AI workloads The listed resources address text and multimodal generation, retrieval-augmented generation, speech recognition and synthesis, image and video processing, coding, structured-data prediction, and physical-world or robotics-oriented applications. ## Casos de uso ### Evaluate models for an AI application Use the model catalog to compare available options by task, provider, publisher, endpoint status, and hardware-related information before selecting a model for an application. ### Build enterprise search and RAG workflows Start with the RAG Blueprint and related retrieval models to connect agents or applications with multimodal enterprise data and authoritative knowledge sources. ### Develop coding and agent workflows Use blueprints and models for coding assistants, agentic reasoning, tool use, or local coding environments, including a DGX Spark coding-assistant example grounded in GPU programming knowledge. ### Create speech, media, and multimodal applications Explore resources for real-time voice agents, multilingual speech recognition, translation, image and video generation, media localization, and lip synchronization. ### Run device-specific or local workloads Follow playbooks for DGX Spark, DGX Station, or RTX Workstations when the goal is to serve models, run inference, fine-tune models, build applications, or operate agents on a supported machine. ## Perguntas frequentes ### What is NVIDIA NIM APIs used for? It is used to explore model endpoints and development resources for building enterprise generative AI applications. The site combines a model catalog with application blueprints and machine-oriented playbooks. ### What kinds of models are available? The catalog includes language, multimodal, vision, speech, image, video, coding, retrieval, and specialized models. Examples in the source include OCR, speech-to-text, translation, semantic search, RAG, image generation, and agentic coding. ### What are NVIDIA blueprints? Blueprints are workflows and code samples intended to help developers build AI applications from the ground up. The collection includes examples for enterprise search, agents, voice, healthcare, finance, retail, media, robotics, and infrastructure. ### Which devices have dedicated playbooks? The playbooks page lists dedicated collections for DGX Spark, DGX Station, and RTX Workstations. The guides are organized around those machines and selected use cases. ### Does the available information specify pricing or complete integration details? No. The supplied pricing page text does not show pricing mechanics, and the provided sources do not establish a complete integration list or universal access terms. Availability and requirements may differ by model, endpoint, or workflow. ## Informações rápidas - **Product type**: Model catalog, API exploration hub, blueprint collection, and playbook library - **Primary audience**: Developers building enterprise generative AI applications - **Model catalog**: 98 models listed in the supplied catalog view - **Blueprint collection**: 33 blueprints listed in the supplied collection view - **Playbook collections**: 48 DGX Spark, 21 DGX Station, and 8 RTX Workstation playbooks listed - **Source domain**: build.nvidia.com