> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fastrun.si/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

> The models fastrun serves, with their limits and features.

## Gemma 4 31B Instruct

```text Model ID theme={null}
google/gemma-4-31b-it
```

Google's Gemma 4 31B, instruction-tuned, with FP8 weights. It runs on vLLM with multi-token-prediction speculative
decoding, which speeds up generation without changing the output distribution.

| | |
| - | - |
| Context window | 131,072 tokens (input and output together) |
| Maximum output | 16,384 tokens |
| Input | Text only |
| Output | Text |
| Quantization | FP8 |
| Region | United States |

### Features

| Feature | Support |
| - | - |
| [Streaming](https://docs.fastrun.si/guides/streaming.md) | Yes |
| [Tool calling](https://docs.fastrun.si/guides/tool-calling.md) | Yes |
| [Structured outputs](https://docs.fastrun.si/guides/structured-outputs.md) (`response_format`) | Yes |
| [Reasoning](https://docs.fastrun.si/guides/reasoning.md) | Yes, off by default |
| [Prompt caching](https://docs.fastrun.si/guides/prompt-caching.md) | Yes, automatic |
| Images, audio, video | No |

### Sampling parameters

`temperature` (0–2), `top_p` (0–1), `top_k` (0–1000), `min_p` (0–1), `frequency_penalty` (−2–2),
`presence_penalty` (−2–2), `repetition_penalty` (0–2), `max_tokens`, `stop` (up to 4), `seed`, `logprobs` and
`top_logprobs` (0–20). See [Create a chat completion](https://docs.fastrun.si/api-reference/chat/create-a-chat-completion.md) for details.

## Listing models

`GET /v1/models` returns the models that can serve requests right now.

```bash theme={null}
curl https://api.fastrun.si/v1/models \
  -H "Authorization: Bearer $FASTRUN_API_KEY"
```

```json theme={null}
{
  "object": "list",
  "data": [
    { "id": "google/gemma-4-31b-it", "object": "model", "owned_by": "fastrun" }
  ]
}
```


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.