AI/EXPLORER
ToolsCategoriesSitesLLMsCompareAI QuizAlternativesPremium
—AI Tools
—Sites & Blogs
—LLMs & Models
—Categories
AI Explorer

Find and compare the best artificial intelligence tools for your projects.

Made within France

Explore

  • ›All tools
  • ›Sites & Blogs
  • ›LLMs & Models
  • ›Compare
  • ›Chatbots
  • ›AI Images
  • ›Code & Dev

Company

  • ›Premium
  • ›About
  • ›Contact
  • ›Blog

Legal

  • ›Legal notice
  • ›Privacy
  • ›Terms

© 2026 AI Explorer·All rights reserved.

HomeLLMsChatLlama 3.2 1B Instruct FP8 dynamic

Llama 3.2 1B Instruct FP8 dynamic

by RedHatAI

Open source · 1M downloads · 4 likes

0.9
(4 reviews)ChatAPI & Local
About

The Llama 3.2 1B Instruct FP8 dynamic model is an optimized and quantized version of the Llama 3.2 1B Instruct model, designed to deliver similar performance while significantly reducing memory and storage requirements. By quantizing both weights and activations to FP8, it halves the disk and GPU memory footprint while maintaining near-identical accuracy to the original. This model is particularly well-suited for English-language conversational assistance tasks, such as chatbots or automated response systems, with enhanced efficiency for deployment on constrained infrastructure. Its key advantages lie in its lightweight design and compatibility with tools like vLLM, while retaining reasoning and comprehension capabilities comparable to the unquantized model.

Documentation

Llama-3.2-1B-Instruct-FP8-dynamic

Model Overview

  • Model Architecture: Meta-Llama-3.2
    • Input: Text
    • Output: Text
  • Model Optimizations:
    • Weight quantization: FP8
    • Activation quantization: FP8
  • Intended Use Cases: Intended for commercial and research use in multiple languages. Similarly to Llama-3.2-1B-Instruct, this models is intended for assistant-like chat.
  • Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in languages other than English.
  • Release Date: 9/25/2024
  • Version: 1.0
  • License(s): llama3.2
  • Model Developers: Neural Magic

Quantized version of Llama-3.2-1B-Instruct. It achieves an average score of 50.88 on a subset of task from the OpenLLM benchmark (version 1), whereas the unquantized model achieves 51.70.

Model Optimizations

This model was obtained by quantizing the weights and activations of Llama-3.2-1B-Instruct to FP8 data type, ready for inference with vLLM built from source. This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%.

Only the weights and activations of the linear operators within transformers blocks are quantized. Symmetric per-channel quantization is applied, in which a linear scaling per output dimension maps the FP8 representations of the quantized weights and activations. Activations are also quantized on a per-token dynamic basis. LLM Compressor is used for quantization.

Deployment

Use with vLLM

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

Python
from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic"

sampling_params = SamplingParams(temperature=0.6, top_p=0.9, 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)

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

This model was created by applying LLM Compressor, as presented in the code snipet below.

Python
import torch

from transformers import AutoTokenizer

from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot
from llmcompressor.transformers.compression.helpers import (  # noqa
    calculate_offload_device_map,
    custom_offload_device_map,
)

recipe = """
quant_stage:
    quant_modifiers:
        QuantizationModifier:
            ignore: ["lm_head"]
            config_groups:
                group_0:
                    weights:
                        num_bits: 8
                        type: float
                        strategy: channel
                        dynamic: false
                        symmetric: true
                    input_activations:
                        num_bits: 8
                        type: float
                        strategy: token
                        dynamic: true
                        symmetric: true
                    targets: ["Linear"]
"""

model_stub = "meta-llama/Llama-3.2-1B-Instruct"
model_name = model_stub.split("/")[-1]

device_map = calculate_offload_device_map(
    model_stub, reserve_for_hessians=False, num_gpus=1, torch_dtype="auto"
)

model = SparseAutoModelForCausalLM.from_pretrained(
    model_stub, torch_dtype="auto", device_map=device_map
)

output_dir = f"./{model_name}-FP8-dynamic"

oneshot(
    model=model,
    recipe=recipe,
    output_dir=output_dir,
    save_compressed=True,
    tokenizer=AutoTokenizer.from_pretrained(model_stub),
)

Evaluation

The model was evaluated on MMLU, ARC-Challenge, GSM-8K, and Winogrande. Evaluation was conducted using the Neural Magic fork of lm-evaluation-harness (branch llama_3.1_instruct) and the vLLM engine. This version of the lm-evaluation-harness includes versions of ARC-Challenge, GSM-8K, MMLU, and MMLU-cot that match the prompting style of Meta-Llama-3.1-Instruct-evals.

Accuracy

Open LLM Leaderboard evaluation scores

Benchmark Llama-3.2-1B-Instruct Llama-3.2-1B-Instruct-FP8-dynamic (this model) Recovery
MMLU (5-shot) 47.66 47.55 99.8%
MMLU-cot (0-shot) 47.10 46.79 99.3%
ARC Challenge (0-shot) 58.36 57.25 98.1%
GSM-8K-cot (8-shot, strict-match) 45.72 45.94 100.5%
Winogrande (5-shot) 62.27 61.40 98.6%
Hellaswag (10-shot) 61.01 60.95 99.9%
TruthfulQA (0-shot, mc2) 43.52 44.23 101.6%
Average 52.24 52.02 99.7%

Reproduction

The results were obtained using the following commands:

MMLU

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,max_model_len=3850,max_gen_toks=10,tensor_parallel_size=1 \
  --tasks mmlu_llama_3.1_instruct \
  --fewshot_as_multiturn \
  --apply_chat_template \
  --num_fewshot 5 \
  --batch_size auto

MMLU-CoT

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,max_model_len=4064,max_gen_toks=1024,tensor_parallel_size=1 \
  --tasks mmlu_cot_0shot_llama_3.1_instruct \
  --apply_chat_template \
  --num_fewshot 0 \
  --batch_size auto

ARC-Challenge

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,max_model_len=3940,max_gen_toks=100,tensor_parallel_size=1 \
  --tasks arc_challenge_llama_3.1_instruct \
  --apply_chat_template \
  --num_fewshot 0 \
  --batch_size auto

GSM-8K

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,max_model_len=4096,max_gen_toks=1024,tensor_parallel_size=1 \
  --tasks gsm8k_cot_llama_3.1_instruct \
  --fewshot_as_multiturn \
  --apply_chat_template \
  --num_fewshot 8 \
  --batch_size auto

Hellaswag

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks hellaswag \
  --num_fewshot 10 \
  --batch_size auto

Winogrande

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks winogrande \
  --num_fewshot 5 \
  --batch_size auto

TruthfulQA

SCSS
lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks truthfulqa \
  --num_fewshot 0 \
  --batch_size auto
Capabilities & Tags
safetensorsllamafp8vllmtext-generationconversationalendefrit
Links & Resources
Specifications
CategoryChat
AccessAPI & Local
LicenseOpen Source
PricingOpen Source
Parameters1B parameters
Rating
0.9

Try Llama 3.2 1B Instruct FP8 dynamic

Access the model directly