furiosa-apps provides end-to-end reference implementations built with Furiosa-LLM, covering benchmarks, interactive demos, and integrations with open-source AI platforms.
- For Furiosa SDK installation and setup instructions, see Quick Start with Furiosa-LLM.
- Clone this repository and navigate the directory, selecting which reference you want to run.
# Clone the repository git clone https://github.com/furiosa-ai/furiosa-apps.git # Navigate to the target reference application cd furiosa-apps/<applications> # Install Python venv python -m venv venv source venv/bin/activate
| Name | Description | Primary Use Case |
|---|---|---|
| Benchmark | A comprehensive evaluation framework for measuring performance and accuracy of LLMs on Furiosa RNGD. | Hardware evaluation and comparison |
| Chat Playground | A full-stack chatbot application with real-time inference metrics, built on Furiosa-LLM. | Interactive LLM inference exploration |
| Coding Agent | A terminal-based AI coding assistant (OpenCode) and a web-based unit-test generator, both powered by Furiosa-LLM. | AI-assisted software development |
| LLM Assistant | A collection of AI assistant applications: an OpenClaw agent platform integration, a multi-agent weekly news reporter, and a real-time financial sentiment analysis dashboard. | Custom AI Support |
| RAG | An end-to-end RAG pipeline using embedding, reranker, and generation models served via Furiosa-LLM. | Knowledge-grounded question answering |
These applications are provided for reference purposes only and are not intended for production use.
The pre-compiled models are available at FuriosaAI models on Hugging Face.
| Model (HF ID) | Required RNGDs | Max Context Length |
|---|---|---|
| furiosa-ai/EXAONE-4.0-32B-FP8 | 4 | 131072 |
| furiosa-ai/Qwen3-32B-FP8 | 4 | 40960 |
| furiosa-ai/Llama-3.3-70B-Instruct | 4 | 131072 |
| furiosa-ai/Llama-3.1-8B-Instruct | 1 | 131072 |
| furiosa-ai/Qwen3-Embedding-8B | 1 | 8192 |
| furiosa-ai/Qwen3-Reranker-8B | 1 | 8192 |
For broader discussions or support of Furiosa SDK and RNGD, we encourage you to participate in the open forum linked below.
This project is licensed under the Apache License 2.0 — see the LICENSE file for details.