***

title: Nitro Variant
subtitle: 'High-speed model inference with :nitro'
headline: Nitro Variant | High-Speed Inference
canonical-url: '[https://openrouter.ai/docs/guides/routing/model-variants/nitro](https://openrouter.ai/docs/guides/routing/model-variants/nitro)'
'og:site\_name': OpenRouter Documentation
'og:title': Nitro Variant - High-Speed Inference
'og:description': 'Access high-speed model inference using the :nitro variant.'
'og:image':
type: url
value: >-
[https://openrouter.ai/dynamic-og?title=Nitro%20Variant\&description=High-speed%20inference](https://openrouter.ai/dynamic-og?title=Nitro%20Variant\&description=High-speed%20inference)
'og:image:width': 1200
'og:image:height': 630
'twitter:card': summary\_large\_image
'twitter:site': '@OpenRouter'
noindex: false
nofollow: false
---------------

The `:nitro` variant is an alias for sorting providers by throughput. When you use `:nitro`, OpenRouter will prioritize providers with the highest throughput (tokens per second).

## Usage

Append `:nitro` to any model ID:

```json
{
  "model": "openai/gpt-5.2:nitro"
}
```

This is exactly equivalent to setting `provider.sort` to `"throughput"` in your request. For more details on provider sorting, see the [Provider Routing documentation](/docs/features/provider-routing#provider-sorting).