---
title: NVIDIA LiteLLM Router
emoji: 🚀
colorFrom: green
colorTo: blue
sdk: docker
pinned: false
license: apache-2.0
---

# NVIDIA LiteLLM Router

Auto-routes across 31 free NVIDIA NIM models with latency-based routing, tier selection, and automatic failover.

## Usage

```bash
curl https://tomoritemitopex-nvidia-litellm-router.hf.space/v1/chat/completions \
  -H "Authorization: Bearer sk-litellm-master" \
  -H "Content-Type: application/json" \
  -d '{"model": "nvidia-auto", "messages": [{"role": "user", "content": "hello"}]}'
```

### Model Groups
| Model | Description |
|-------|-------------|
| `nvidia-auto` | Fastest across ALL models |
| `nvidia-coding` | Fastest coding model |
| `nvidia-reasoning` | Fastest reasoning model |
| `nvidia-general` | Fastest general model |
| `nvidia-fast` | Fastest small/efficient model |
| `<model-name>` | Direct access (e.g. `deepseek-v3.2`, `kimi-k2-instruct`) |

### Features
- Latency-based routing picks the fastest model automatically
- 429 rate limit? Retries 3x with backoff, then failover
- Model slow? Deprioritized automatically
- Model down? 60s cooldown, auto-recovers