NVIDIA NIM
Researched 24d agoInference PlatformTier 2NVIDIA
NVIDIA NIM offers 89 tracked models (1 with output token pricing). This catalog covers coding, rag, and agents; open any model detail page for benchmarks, batch tiers, and migration prompts. 54 retired models are listed below.
Covers 6 workload areas across 89 available models and 54 retired; last verified 2026-10-26.
Use it for
- Teams comparing token and batch pricing across this provider's models
- Operators routing coding, rag, and agents workloads through this API
Do not use it for
- Final benchmark picks without opening the relevant model detail page
Tracked models
89
Models available through this provider
Priced output routes
1
Models with output token pricing tracked
Cheapest output
$0.500
Nemotron 3 Super-120B-A12B on this route
Batch-ready models
0
No batch pricing tracked
Latest model release
2026-05-31
172d since newest release
Freshness
2026-10-26
Researched 24d ago
Routes available via routers & gateways
Information
NVIDIA NIM is NVIDIA's deployment platform for GPU-accelerated inference microservices. Developers can try hosted NIM APIs through the NVIDIA API Catalog on build.nvidia.com, then move the same model families into self-hosted NIM containers on NVIDIA GPUs in a data center, private cloud, public cloud, or workstation. The catalog positions NIM around optimized open and NVIDIA models, including chat, coding, reasoning, retrieval, vision, speech, and safety use cases, with downloadable model cards and API endpoints where NVIDIA exposes them.
Where this host wins
- Coding: 1 tracked model with SWE-bench / HumanEval-style scores.
- RAG: 1 tracked model with ruler / needle retrieval benchmarks.
- Agentic: 1 tracked model with BFCL, tau-bench, and SWE-bench tool-use coverage.
- Long-context: 1 tracked model with context-token or InfiniteBench-class signal.
Getting started
Platform Overview
NIM packages inference runtimes and model profiles into containers that expose standard API surfaces such as chat completions, completions, model listing, tokenization, health, and management endpoints. The hosted API path is useful for prototyping and catalog discovery, while the NGC/container path is the self-hosted route for teams that want GPU-hour infrastructure control, private-network deployment, Kubernetes scaling, or NVIDIA AI Enterprise support.
Available Models(89)
View all →| Model | Input (per 1M) | Output (per 1M) | Type |
|---|---|---|---|
| — | — | Provisioned | |
| — | — | Provisioned | |
| — | — | Provisioned | |
| — | — | Serverless | |
| $0.10 | $0.50 | Serverless | |
| — | — | Serverless | |
| — | — | Serverless | |
| — | — | Serverless | |
| — | — | Serverless | |
| — | — | Serverless |
Retired models(54)
View all →| Model | Type |
|---|---|
Kimi K2.6Deprecated Scheduled retirement: December 31, 2026 | Serverless |
Kimi K2.5Deprecated | Serverless |
Mistral Medium 3Deprecated Retired May 11, 2026 | Serverless |
Gemma 3n 4B (free)Deprecated | Serverless |
Nemotron 4 340BDeprecated | Provisioned |
DeepSeek R1Deprecated Retired January 26, 2026 | Serverless |
Gemma 2 9B SahabatAI InstructDeprecated Retired April 15, 2026 | Serverless |
DeepSeek V3Deprecated | Serverless |
Yi LargeDeprecated | Provisioned |
Llama 3 ChatQA 70BDeprecated | Provisioned |