Llama-3.3-70B-Instruct-FP8-dynamic

Llama-3.3-70B-Instruct-FP8-dynamic

This repository contains RedHatAI/Llama-3.3-70B-Instruct-FP8-dynamic together with a Furiosa Executable Bundle (FXB) for running it on FuriosaAI RNGD with Furiosa-LLM. The same model also runs on other frameworks (such as vLLM, SGLang, and Transformers); for usage with those, see the upstream RedHatAI/Llama-3.3-70B-Instruct-FP8-dynamic model card.

Overview

Llama-3.3-70B-Instruct is Meta's 70B instruction-tuned model, an auto-regressive dense transformer optimized for multilingual dialogue, instruction following, and tool usage, with quality competitive with much larger models. This is the FP8-dynamic quantized variant published by RedHatAI, derived from meta-llama/Llama-3.3-70B-Instruct. Its intended use is the same as the upstream RedHatAI/Llama-3.3-70B-Instruct-FP8-dynamic, and it is released under the Llama 3.3 Community License.

• Architecture: Llama 3.3 (dense)
• Input / Output: Text / Text
• Supported Inference Engine: Furiosa LLM
• Supported Hardware: FuriosaAI RNGD

Quantization

The linear (Transformer block) weights are quantized to FP8 (static), and activations use dynamic FP8 quantization at runtime. Activations are quantized per token at inference time rather than with offline calibration.

Features

• Tool calling. The model supports tool (function) calling through the llama3_json tool-call parser, the parser used by the Llama 3 series.

Parallelism Strategy

On RNGD, Llama-3.3-70B-Instruct-FP8-dynamic runs with a tensor-parallel size of 32 PEs, which maps to four RNGD cards (8 PEs per card).

Usage

To run this model with Furiosa-LLM, follow the example commands below after installing Furiosa-LLM and its prerequisites.

Launch the server

The simplest way to serve the model is:

# Launch the server, listening on port 8000 by default
furiosa-llm serve furiosa-ai/Llama-3.3-70B-Instruct-FP8-dynamic

To also enable tool (function) calling, add the llama3_json tool-call parser (the parser used by the Llama 3 series):

furiosa-llm serve furiosa-ai/Llama-3.3-70B-Instruct-FP8-dynamic \
  --enable-auto-tool-choice \
  --tool-call-parser llama3_json

When the server is ready, you will see:

INFO:     Started server process [27507]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

Basic Usage

The server exposes an OpenAI-compatible API. You can send a request with curl:

curl http://localhost:8000/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{
    "model": "furiosa-ai/Llama-3.3-70B-Instruct-FP8-dynamic",
    "messages": [{"role": "user", "content": "What is the capital of France?"}]
    }' \
    | python -m json.tool

Advanced Usage

Tool calling. With the server launched using --enable-auto-tool-choice --tool-call-parser llama3_json (see Launch the server), pass tools in the request and let the model decide when to call them. See the Tool Calling guide for a complete client example and details on tool-choice options.

Learn more

• Tool Calling — parsers, tool-choice options, and more examples
• Furiosa-LLM Server (furiosa-llm serve) — full OpenAI-compatible API reference and serving options
• RedHatAI/Llama-3.3-70B-Instruct-FP8-dynamic — upstream model card