Nemotron models
NVIDIA publishes 67 models under its own name in the API catalogue, including the Nemotron family. Nemotron 3.5 Super VL 120B A12B is described as an early-access reasoning model for agentic, coding, tool-calling and multimodal workloads, while nemotron-3.5-lightning-30b-a3b is a 30B A3B mixture-of-experts model aimed at specialised agentic tasks. The retrieval line includes nemotron-3.5-parse for extracting text, tables and document layout from page images, alongside embedding and reranking models for RAG applications.
- Nemotron 3.5 Super VL 120B A12B — multimodal reasoning, early access
- nemotron-3.5-lightning-30b-a3b — 30B A3B MoE for agentic tasks
- nemotron-3.5-parse — document text, table and layout extraction
- nemotron-3.5-rerank-8b and nemotron-3.5-embed-1b/8b — retrieval models
- nemotron-3-ultra-550b-a55b — long-context MoE reasoning model
Partner and open models
The catalogue also lists models contributed by other organisations and served through NIM endpoints. Mistral Large 4 is a 1T-total, 52B-active granular mixture-of-experts model accepting text and image input, with reasoning, function calling and structured outputs, served in NVFP4 with DSpark speculative decoding. OpenAI entries include gpt-oss-20b, a smaller mixture-of-experts text-only model for reasoning and maths, and whisper-large-v3 for speech recognition. Black Forest Labs supplies the FLUX image models, among them FLUX.1-schnell, a distilled image generation model.
- Mistral Large 4 — MoE text and image model with tool use
- gpt-oss-20b — text-only MoE for reasoning and maths
- whisper-large-v3 — multilingual speech recognition, available as a NIM
- FLUX.1-schnell — distilled fast image generation
- FLUX.1-Kontext-dev — in-context image generation and editing
NemoClaw and agentic tooling
NemoClaw is presented on the build.nvidia.com home page under the theme of safe agent execution, access control and data protection. It appears both as a deployable agent and as the basis of several blueprints, including NemoClaw for Agent Optimization, NemoClaw for LangChain Deep Agents Code, NemoClaw for Hermes Agent and NemoClaw for OpenClaw. DGX Station and DGX Spark playbooks show how to run NemoClaw with a local LLM inside an OpenShell sandbox using vLLM inference, with optional Telegram integration.
- Safe agent execution with access control
- NemoClaw blueprints for LangChain, Hermes and OpenClaw
- Local deployment playbook with vLLM and OpenShell
Blueprints
Blueprints are workflows and code samples for building AI applications from the ground up, with 36 listed under NVIDIA in the catalogue. Build a RAG Pipeline Blueprint is tagged with NVIDIA AI, NIM, Nemotron, retrieval-augmented generation and NeMo Retriever, and is offered as a launchable developer example. Other blueprints cover documentation agents, life sciences, deep research, AI factory operations and GPU query engines.
- Build a RAG Pipeline Blueprint — NeMo Retriever and NIM based
- DocsFab — documentation agent
- Agent For Life Sciences — BioNeMo and drug discovery
- NVIDIA Deep Researcher Agent
- AI Factory Operations Agent
- GPU Query Engine (GQE)
DGX Spark and DGX Station playbooks
DGX Spark has its own section of the catalogue offering playbooks for 64 GB and 128 GB configurations, with step-by-step instructions for common AI workflows. Topics include connecting multiple DGX Sparks for distributed workloads, serving LLMs with vLLM on one device, chatting with LLMs via Open WebUI and Ollama, fine-tuning vision-language models and building knowledge graphs with txt2kg. DGX Station playbooks target the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip and cover agent skills, ComfyUI image and video generation, and fine-tuning Isaac GR00T for robot skills.
- Playbooks for DGX Spark 64 GB and 128 GB
- Serve LLMs with vLLM on one device or across nodes
- Generate images and videos with ComfyUI
- Fine-tune Isaac GR00T for robot skills
- Build knowledge graphs with txt2kg
Sources
-
AI Models by Black_forest_labs | Try NVIDIA NIM APIs https://build.nvidia.com/black-forest-labs Verified 03 Nov 2026
-
AI Models by Mistral AI | Try NVIDIA NIM APIs https://build.nvidia.com/mistralai Verified 03 Nov 2026
-
Explore Run on RTX Workstations | Try NVIDIA NIM APIs https://build.nvidia.com/explore/run-on-rtx Verified 03 Nov 2026
-
Explore Reasoning Models | Try NVIDIA NIM APIs https://build.nvidia.com/explore/reasoning Verified 03 Nov 2026
-
Explore Vision Models | Try NVIDIA NIM APIs https://build.nvidia.com/explore/vision Verified 03 Nov 2026
-
Explore Retrieval Models | Try NVIDIA NIM APIs https://build.nvidia.com/explore/retrieval Verified 03 Nov 2026
-
Explore Biology Models | Try NVIDIA NIM APIs https://build.nvidia.com/explore/biology Verified 03 Nov 2026
Last verified 03 Nov 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.