Qwen3-32B-quantized.w4a16

Qwen3-32B-quantized.w4a16

Model Overview

• Model Architecture: Qwen3ForCausalLM
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: INT4
• Intended Use Cases:
  • Reasoning.
  • Function calling.
  • Subject matter experts via fine-tuning.
  • Multilingual instruction following.
  • Translation.
• Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws).
• Release Date: 05/05/2025
• Version: 1.0
• Model Developers: RedHat (Neural Magic)

Model Optimizations

This model was obtained by quantizing the weights of Qwen3-32B to INT4 data type. 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 of the linear operators within transformers blocks are quantized. Weights are quantized using a symmetric per-group scheme, with group size 128. The GPTQ algorithm is applied for quantization, as implemented in the llm-compressor library.

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/Qwen3-32B-quantized.w4a16"
number_gpus = 1
sampling_params = SamplingParams(temperature=0.6, top_p=0.95, top_k=20, min_p=0, max_tokens=256)

messages = [
    {"role": "user", "content": prompt}
]

tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [{"role": "user", "content": "Give me a short introduction to large language model."}]

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

<details> <summary>Creation details</summary> This model was created with [llm-compressor](https://github.com/vllm-project/llm-compressor) by running the code snippet below.

from llmcompressor.modifiers.quantization import GPTQModifier
from llmcompressor.transformers import oneshot
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model
model_stub = "Qwen/Qwen3-32B"
model_name = model_stub.split("/")[-1]

num_samples = 1024
max_seq_len = 8192

model = AutoModelForCausalLM.from_pretrained(model_stub)

tokenizer = AutoTokenizer.from_pretrained(model_stub)

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

ds = load_dataset("neuralmagic/LLM_compression_calibration", split="train")
ds = ds.map(preprocess_fn)

# Configure the quantization algorithm and scheme
recipe = GPTQModifier(
    ignore=["lm_head"],
    sequential_targets=["Qwen3DecoderLayer"],
    targets="Linear",
    scheme="W4A16",
    dampening_frac=0.1,
)

# Apply quantization
oneshot(
    model=model,
    dataset=ds, 
    recipe=recipe,
    max_seq_length=max_seq_len,
    num_calibration_samples=num_samples,
)

# Save to disk in compressed-tensors format
save_path = model_name + "-quantized.w4a16"
model.save_pretrained(save_path)
tokenizer.save_pretrained(save_path)
print(f"Model and tokenizer saved to: {save_path}")

</details>

Evaluation

The model was evaluated on the OpenLLM leaderboard tasks (versions 1 and 2), using lm-evaluation-harness, and on reasoning tasks using lighteval. vLLM was used for all evaluations.

<details> <summary>Evaluation details</summary>

lm-evaluation-harness

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-32B-quantized.w4a16",dtype=auto,gpu_memory_utilization=0.5,max_model_len=8192,enable_chunk_prefill=True,tensor_parallel_size=1 \
  --tasks openllm \
  --apply_chat_template\
  --fewshot_as_multiturn \
  --batch_size auto

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-32B-quantized.w4a16",dtype=auto,gpu_memory_utilization=0.5,max_model_len=8192,enable_chunk_prefill=True,tensor_parallel_size=1 \
  --tasks mgsm \
  --apply_chat_template\
  --batch_size auto

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-32B-quantized.w4a16",dtype=auto,gpu_memory_utilization=0.5,max_model_len=16384,enable_chunk_prefill=True,tensor_parallel_size=1 \
  --tasks leaderboard \
  --apply_chat_template\
  --fewshot_as_multiturn \
  --batch_size auto

lighteval

lighteval_model_arguments.yaml

model_parameters:
  model_name: RedHatAI/Qwen3-32B-quantized.w4a16
  dtype: auto
  gpu_memory_utilization: 0.9
  max_model_length: 40960
  generation_parameters:
    temperature: 0.6
    top_k: 20
    min_p: 0.0
    top_p: 0.95
    max_new_tokens: 32768

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|aime24|0|0 \
  --use_chat_template = true

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|aime25|0|0 \
  --use_chat_template = true

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|math_500|0|0 \
  --use_chat_template = true

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks lighteval|gpqa:diamond|0|0 \
  --use_chat_template = true

lighteval vllm \
  --model_args lighteval_model_arguments.yaml \
  --tasks extended|lcb:codegeneration \
  --use_chat_template = true

</details>

Accuracy

<table> <tr> <th>Category </th> <th>Benchmark </th> <th>Qwen3-32B </th> <th>Qwen3-32B-quantized.w4a16<br>(this model) </th> <th>Recovery </th> </tr> <tr> <td rowspan="7" ><strong>OpenLLM v1</strong> </td> <td>MMLU (5-shot) </td> <td>80.96 </td> <td>80.36 </td> <td>99.3% </td> </tr> <tr> <td>ARC Challenge (25-shot) </td> <td>69.03 </td> <td>68.69 </td> <td>99.5% </td> </tr> <tr> <td>GSM-8K (5-shot, strict-match) </td> <td>87.64 </td> <td>85.97 </td> <td>98.1% </td> </tr> <tr> <td>Hellaswag (10-shot) </td> <td>71.10 </td> <td>71.18 </td> <td>100.1% </td> </tr> <tr> <td>Winogrande (5-shot) </td> <td>69.77 </td> <td>70.90 </td> <td>100.5% </td> </tr> <tr> <td>TruthfulQA (0-shot, mc2) </td> <td>58.63 </td> <td>58.86 </td> <td>100.4% </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>72.86</strong> </td> <td><strong>72.52</strong> </td> <td><strong>99.6%</strong> </td> </tr> <tr> <td rowspan="7" ><strong>OpenLLM v2</strong> </td> <td>MMLU-Pro (5-shot) </td> <td>54.24 </td> <td>52.63 </td> <td>97.03% </td> </tr> <tr> <td>IFEval (0-shot) </td> <td>86.23 </td> <td>85.53 </td> <td>99.2% </td> </tr> <tr> <td>BBH (3-shot) </td> <td>44.29 </td> <td>41.07 </td> <td>92.7% </td> </tr> <tr> <td>Math-lvl-5 (4-shot) </td> <td>54.61 </td> <td>55.38 </td> <td>101.4% </td> </tr> <tr> <td>GPQA (0-shot) </td> <td>5.53 </td> <td>4.59 </td> <td>--- </td> </tr> <tr> <td>MuSR (0-shot) </td> <td>7.85 </td> <td>8.25 </td> <td>--- </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>42.13</strong> </td> <td><strong>41.24</strong> </td> <td><strong>97.9%</strong> </td> </tr> <tr> <td><strong>Multilingual</strong> </td> <td>MGSM (0-shot) </td> <td>32.57 </td> <td>33.77 </td> <td>103.7% </td> </tr> <tr> <td rowspan="6" ><strong>Reasoning<br>(generation)</strong> </td> <td>AIME 2024 </td> <td>79.37 </td> <td>77.29 </td> <td>97.4% </td> </tr> <tr> <td>AIME 2025 </td> <td>71.77 </td> <td>64.27 </td> <td>89.6% </td> </tr> <tr> <td>GPQA diamond </td> <td>66.67 </td> <td>66.67 </td> <td>100.0% </td> </tr> <tr> <td>Math-lvl-5 </td> <td>96.20 </td> <td>97.20 </td> <td>101.0% </td> </tr> <tr> <td>LiveCodeBench </td> <td>62.45 </td> <td>59.63 </td> <td>95.5% </td> </tr> </table>