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

# Fireworks AI

> Use Fireworks AI models in your BoostGPT AI teammates

## Overview

Fireworks serves open-weight models — Llama, Mixtral, Qwen, DeepSeek and Yi — behind one fast inference endpoint, so you can pick an open model without hosting it yourself.

## Available Models

> [card] DeepSeek V3
>
> **4 credits** • Latest DeepSeek model. Improved performance and capabilities.
>
> * 64K context window
> * Good reasoning
> * Speed: Fast • Cost: Low
> * Model id: `deepseek-v3`

> [card] DeepSeek R1
>
> **25 credits** • Specialized reasoning model. Enhanced analytical capabilities.
>
> * 64K context window
> * Excellent reasoning (reasoning model)
> * Speed: Medium • Cost: Low
> * Model id: `deepseek-r1`

> [card] Llama V3 (8B)
>
> **2 credits** • Instruction-tuned 8B model. Efficient for general tasks.
>
> * 8K context window
> * Good reasoning
> * Speed: Very\_fast • Cost: Very\_low
> * Model id: `llama-v3-8b-instruct`

> [card] Llama V3 (70B)
>
> **8 credits** • Large instruction-tuned model. Powerful for complex tasks.
>
> * 8K context window
> * Excellent reasoning
> * Speed: Medium • Cost: Medium
> * Model id: `llama-v3-70b-instruct`

> [card] Llama V3.1 (8B)
>
> **2 credits** • Updated 8B instruction model. Improved performance and efficiency.
>
> * 128K context window
> * Good reasoning
> * Speed: Very\_fast • Cost: Very\_low
> * Model id: `llama-v3p1-8b-instruct`

> [card] Llama V3.1 (70B)
>
> **8 credits** • Updated 70B instruction model. Enhanced capabilities and performance.
>
> * 128K context window
> * Excellent reasoning
> * Speed: Fast • Cost: Low
> * Model id: `llama-v3p1-70b-instruct`

> [card] Llama V3.1 (405B)
>
> **8 credits** • Massive 405B parameter model. Exceptional performance on complex tasks.
>
> * 128K context window
> * Exceptional reasoning
> * Speed: Slow • Cost: High
> * Model id: `llama-v3p1-405b-instruct`

> [card] Yi Large
>
> **8 credits** • Large Yi model. Strong performance across diverse applications.
>
> * 32K context window
> * Excellent reasoning
> * Speed: Medium • Cost: Medium
> * Model id: `yi-large`

> [card] Mixtral (8x7B)
>
> **4 credits** • Instruction-tuned Mixtral model. Efficient mixture of experts architecture.
>
> * 32K context window
> * Good reasoning
> * Speed: Fast • Cost: Low
> * Model id: `mixtral-8x7b-instruct`

> [card] Mixtral (8x22B)
>
> **10 credits** • Larger instruction-tuned Mixtral. Enhanced capabilities with larger capacity.
>
> * 64K context window
> * Excellent reasoning
> * Speed: Medium • Cost: Medium
> * Model id: `mixtral-8x22b-instruct`

> [card] Qwen 2.5 (72B)
>
> **8 credits** • Large Qwen model. Advanced capabilities for complex tasks.
>
> * 128K context window
> * Excellent reasoning
> * Speed: Medium • Cost: Medium
> * Model id: `qwen2p5-72b-instruct`

## Setup

### Using BoostGPT-Hosted API Keys

**1. Select a model**

In your BoostGPT dashboard, choose any Fireworks AI model when creating or configuring an AI teammate.

**2. Start chatting**

Credits are deducted per step at the rate shown on each model above. Nothing else to configure.

### Using Your Own API Key

Bring your own key to bill Fireworks AI directly instead of spending BoostGPT credits. See
[Bring Your Own Keys](/docs/providers/bring-your-own-keys.md).

```javascript theme={null}
await client.chat({
  bot_id: 'bot-id',
  message: 'Hello',
  model: 'deepseek-v3',
  provider_key: process.env.FIREWORKS_API_KEY
});
```

## Next Steps

> [card] [Model Comparison](/docs/providers/model-comparison.md)
>
> Compare every model across all providers

> [card] [Provider Overview](/docs/providers/overview.md)
>
> See all supported providers


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