NVIDIA NIM for Vision Language Models (VLMs)

About NVIDIA NIM for VLMs

• Overview
• Release Notes

Get Started

• Get Started with NIM
• Query the Ministral 3 14B Instruct 2512 API
• Query the Qwen3.5-397B-A17B API
• Query the Kimi-K2.5 API
• Query the API (Past Releases)
  • Mistral Large 3 675B Instruct 2512
  • Cosmos Reason2
  • Nemotron Nano 12B v2 VL
  • Nemotron Parse
  • Cosmos Reason1 7B
  • Llama 4 Maverick 17B 128E Instruct
  • Llama 4 Scout 17B 16E Instruct
  • Mistral Small 3.2 24B Instruct 2506
  • Llama 3.1 Nemotron Nano VL 8B v1
  • Llama 3.2 Vision
  • nemoretriever-parse

Deploy NIM

• Deploy with Helm
• Air Gap Deployment

Work with Models

• Support Matrix
• Ministral 3 14B Instruct 2512 Model Card
• Qwen3.5-397B-A17B Model Card
• Kimi-K2.5 Model Card
• Model Cards (Past Releases)
  • Mistral Large 3 675B Instruct 2512 Model Card
  • Cosmos Reason2 2B Model Card
  • Cosmos Reason2 8B Model Card
  • Nemotron Nano 12B v2 VL Model Card
  • Nemotron Parse Overview
  • Cosmos Reason1 7B Model Card
  • Llama 4 Model Card on GitHub
  • Mistral Small 3.2 24B Instruct 2506 Model Card
  • Llama Nemotron Nano VL Model Card
  • Llama 3.2 Vision Model Card
  • nemoretriever-parse Overview
• Model Profiles
• Fine-Tune a Model

Use Key Features

• Call Functions (Tools)
• Observability
• Structured Generation

Configure Your NIM

• Configure Your NIM
• Benchmarking
• KV Cache Reuse (a.k.a. prefix caching)
• Use OCI S3 Object Storage

Reference

• API Reference
• Utilities
• Sampling Control
• Notes on NIM Container Variants

Notices

• Acknowledgements
• EULA