Llama-3.3-70B-Instruct-FP8-dynamic 👁 Model Icon

👁 Validated Badge

Model Overview

• Model Architecture: Meta-Llama-3.1
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Intended Use Cases: Intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.3 model also supports the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. The Llama 3.3 Community License allows for these use cases.
• Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in any other way that is prohibited by the Acceptable Use Policy and Llama 3.3 Community License. Use in languages beyond English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
• Release Date: 12/11/2024
• Version: 1.0
• Validated on: RHOAI 2.20, RHAIIS 3.0, RHELAI 1.5
• License(s): llama3.3
• Model Developers: RedHat (Neural Magic)

Model Optimizations

This model was obtained by quantizing activation and weights of Llama-3.3-70B-Instruct to FP8 data type. This optimization reduces the number of bits used to represent weights and activations from 16 to 8, reducing GPU memory requirements (by approximately 50%) and increasing matrix-multiply compute throughput (by approximately 2x). Weight quantization also reduces disk size requirements by approximately 50%.

Only weights and activations of the linear operators within transformers blocks are quantized. Weights are quantized with a symmetric static per-channel scheme, whereas activations are quantized with a symmetric dynamic per-token scheme. The llm-compressor library is used for quantization.

Deployment

This model can be deployed efficiently using the vLLM backend, as shown in the example below.

from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "RedHatAI/Llama-3.3-70B-Instruct-FP8-dynamic"
number_gpus = 1

sampling_params = SamplingParams(temperature=0.7, top_p=0.8, max_tokens=256)

tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [
 {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
 {"role": "user", "content": "Who are you?"},
]

prompts = tokenizer.apply_chat_template(messages, tokenize=False)

llm = LLM(model=model_id, tensor_parallel_size=number_gpus)

outputs = llm.generate(prompts, sampling_params)

generated_text = outputs[0].outputs[0].text
print(generated_text)

vLLM aslo supports OpenAI-compatible serving. See the documentation for more details.

Creation

Evaluation

This model was evaluated on the well-known OpenLLM v1, OpenLLM v2, HumanEval, and HumanEval+ benchmarks. In all cases, model outputs were generated with the vLLM engine.

OpenLLM v1 and v2 evaluations were conducted using lm-evaluation-harness and the prompting style of Meta-Llama-3.1-Instruct-evals when available.

HumanEval and HumanEval+ evaluations were conducted using Neural Magic's fork of the EvalPlus repository.

Accuracy

Category	Benchmark	Llama-3.3-70B-Instruct	Llama-3.3-70B-Instruct-FP8-dynamic (this model)	Recovery
OpenLLM v1	MMLU (5-shot)	81.60	81.31	99.6%
	MMLU (CoT, 0-shot)	86.58	86.34	99.7%
	ARC Challenge (0-shot)	49.23	51.96	105.6%
	GSM-8K (CoT, 8-shot, strict-match)	94.16	94.92	100.8%
	Hellaswag (10-shot)	86.49	86.43	99.9%
	Winogrande (5-shot)	84.77	84.53	99.7%
	TruthfulQA (0-shot, mc2)	62.75	63.21	100.7%
	Average	77.94	78.39	100.6%
OpenLLM v2	MMLU-Pro (5-shot)	51.89	51.50	99.3%
	IFEval (0-shot)	90.89	90.92	100.0%
	BBH (3-shot)	63.15	62.84	99.5%
	Math-lvl-5 (4-shot)	0.17	0.33	N/A
	GPQA (0-shot)	46.10	46.30	100.4%
	MuSR (0-shot)	44.35	43.96	99.1%
	Average	49.42	49.31	99.8%
Coding	HumanEval pass@1	83.20	83.70	100.6%
	HumanEval+ pass@1	78.40	78.70	100.4%
Multilingual	Portuguese MMLU (5-shot)	79.76	79.75	100.0%
	Spanish MMLU (5-shot)	79.33	79.17	99.8%
	Italian MMLU (5-shot)	79.15	78.84	99.6%
	German MMLU (5-shot)	77.94	77.95	100.0%
	French MMLU (5-shot)	75.69	75.45	99.7%
	Hindi MMLU (5-shot)	73.81	73.71	99.9%
	Thai MMLU (5-shot)	71.98	71.77	99.7%

Base model

meta-llama/Llama-3.1-70B

meta-llama/Llama-3.3-70B-Instruct

(150)

Quantized variants of Meta Llama 3.3 multilingual large language model (LLM) is an instruction tuned generative model in 70B (text in/text out). • 3 items • Updated Apr 30 • 1

May 2025 Collection of third-party generative AI models validated by Red Hat AI for use across the Red Hat AI Product Portfolio. • 39 items • Updated Apr 30 • 20