Model Architecture: Qwen/Qwen3-32B
- Input: Text
- Output: Text
Model Optimizations:
- Weight quantization: FP4
- Activation quantization: FP4
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: 6/25/2025
Version: 1.0
Model Developers: RedHatAI

This model is a quantized version of Qwen/Qwen3-32B. It was evaluated on a several tasks to assess the its quality in comparison to the unquatized model.

Model Optimizations

This model was obtained by quantizing the weights and activations of Qwen/Qwen3-32B to FP4 data type, ready for inference with vLLM>=0.9.1 This optimization reduces the number of bits per parameter from 16 to 4, reducing the disk size and GPU memory requirements by approximately 75%.

Only the weights and activations of the linear operators within transformers blocks are quantized using LLM Compressor.

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 = "RedHatAI/Qwen3-32B-NVFP4"
number_gpus = 2

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, add_generation_prompt=True, 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

This model was created by applying LLM Compressor with calibration samples from UltraChat, as presented in the code snipet below.

python

from datasets import load_dataset
from transformers import AutoModelForCausalLM, AutoTokenizer

from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.utils import dispatch_for_generation

MODEL_ID = "Qwen/Qwen3-32B"

# Load model.
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

DATASET_ID = "HuggingFaceH4/ultrachat_200k"
DATASET_SPLIT = "train_sft"

# Select number of samples. 512 samples is a good place to start.
# Increasing the number of samples can improve accuracy.
NUM_CALIBRATION_SAMPLES = 512
MAX_SEQUENCE_LENGTH = 2048

# Load dataset and preprocess.
ds = load_dataset(DATASET_ID, split=f"{DATASET_SPLIT}[:{NUM_CALIBRATION_SAMPLES}]")
ds = ds.shuffle(seed=42)

def preprocess(example):
    return {
        "text": tokenizer.apply_chat_template(
            example["messages"],
            tokenize=False,
        )
    }

ds = ds.map(preprocess)

# Tokenize inputs.
def tokenize(sample):
    return tokenizer(
        sample["text"],
        padding=False,
        max_length=MAX_SEQUENCE_LENGTH,
        truncation=True,
        add_special_tokens=False,
    )

ds = ds.map(tokenize, remove_columns=ds.column_names)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp4 with per group 16 via ptq
#   * calibrate a global_scale for activations, which will be used to
#       quantize activations to fp4 on the fly
recipe = QuantizationModifier(targets="Linear", scheme="NVFP4", ignore=["lm_head"])

# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-NVFP4"

# Apply quantization.
oneshot(
    model=model,
    dataset=ds,
    recipe=recipe,
    max_seq_length=MAX_SEQUENCE_LENGTH,
    num_calibration_samples=NUM_CALIBRATION_SAMPLES,
    output_dir=SAVE_DIR,
)

print("\n\n")
print("========== SAMPLE GENERATION ==============")
dispatch_for_generation(model)
input_ids = tokenizer("Hello my name is", return_tensors="pt").input_ids.to("cuda")
output = model.generate(input_ids, max_new_tokens=100)
print(tokenizer.decode(output[0]))
print("==========================================\n\n")

model.save_pretrained(SAVE_DIR, save_compressed=True)
tokenizer.save_pretrained(SAVE_DIR)

Evaluation

This model was evaluated on the well-known OpenLLM v1, OpenLLM v2 and HumanEval_64 benchmarks using lm-evaluation-harness. The Reasoning evals were done using ligheval.

Accuracy

Category	Metric	Qwen/Qwen3-32B	RedHatAI/Qwen3-32B-NVFP4 (this model)	Recovery (%)
OpenLLM V1	arc_challenge	70.65	70.22	99.39
gsm8k	74.15	74.68	100.71
hellaswag	84.00	83.33	99.20
mmlu	81.84	81.23	99.25
truthfulqa_mc2	59.36	58.92	99.26
winogrande	75.93	76.80	101.15
Average	74.32	74.20	99.83
OpenLLM V2	BBH (3-shot)	62.35	60.72	97.39
MMLU-Pro (5-shot)	54.39	51.13	94.01
MuSR (0-shot)	39.29	41.01	104.38
IFEval (0-shot)	88.97	87.29	98.11
GPQA (0-shot)	30.12	30.29	100.56
Math-\|v\|-5 (4-shot)	58.99	56.27	95.39
Average	55.69	54.45	97.79
Coding	HumanEval_64 pass@2	90.14	90.40	100.29
Reasoning	AIME24 (0-shot)	75.86	68.97	90.93
AIME25 (0-shot)	72.41	65.52	90.52
GPQA (Diamond, 0-shot)	62.94	64.47	102.43
Average	70.40	66.32	94.21

Reproduction

The results were obtained using the following commands:

OpenLLM v1

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-32B-NVFP4",dtype=auto,max_model_len=4096,tensor_parallel_size=2,enable_chunked_prefill=True,enforce_eager=True\
  --apply_chat_template \
  --fewshot_as_multiturn \
  --tasks openllm \
  --batch_size auto

OpenLLM v2

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-32B-NVFP4",dtype=auto,max_model_len=4096,tensor_parallel_size=2,enable_chunked_prefill=True,enforce_eager=True\
  --apply_chat_template \
  --fewshot_as_multiturn \
  --tasks leaderboard \
  --batch_size auto

HumanEval_64

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-32B-NVFP4",dtype=auto,max_model_len=4096,tensor_parallel_size=2,enable_chunked_prefill=True,enforce_eager=True\
  --apply_chat_template \
  --fewshot_as_multiturn \
  --tasks humaneval_64_instruct \
  --batch_size auto

LightEval

# --- model_args.yaml ---
cat > model_args.yaml <<'YAML'
model_parameters:
  model_name: "RedHatAI/Qwen3-32B-NVFP4"
  dtype: auto
  gpu_memory_utilization: 0.9
  tensor_parallel_size: 2
  max_model_length: 40960
  generation_parameters:
    seed: 42
    temperature: 0.6
    top_k: 20
    top_p: 0.95
    min_p: 0.0
    max_new_tokens: 32768
YAML

lighteval vllm model_args.yaml \
  "lighteval|aime24|0,lighteval|aime25|0,lighteval|gpqa:diamond|0" \
  --max-samples -1 \
  --output-dir out_dir