Model Architecture: Qwen3MoeForCausalLM
- Input: Text
- Output: Text
Model Optimizations:
- Weight quantization: FP8
- Activation quantization: FP8
Release Date:
Version: 1.0
Model Developers:: Red Hat

Quantized version of Qwen/Qwen3-30B-A3B.

Model Optimizations

This model was obtained by quantizing the weights and activations of Qwen/Qwen3-30B-A3B to FP8 data type. 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 of the language model are quantized.

Deployment

Use with vLLM

Initialize vLLM server:

vllm serve RedHatAI/Qwen3-30B-A3B-FP8-block --tensor_parallel_size 4

Send requests to the server:

python

from openai import OpenAI

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://<your-server-host>:8000/v1"

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

model = "RedHatAI/Qwen3-30B-A3B-FP8-block"

messages = [
    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
]

outputs = client.chat.completions.create(
    model=model,
    messages=messages,
)

generated_text = outputs.choices[0].message.content
print(generated_text)

Creation

This model was quantized using the llm-compressor library as shown below.

Creation details

python

from transformers import AutoProcessor, Qwen3MoeForCausalLM

from llmcompressor import oneshot
from llmcompressor.modeling import replace_modules_for_calibration
from llmcompressor.modifiers.quantization import QuantizationModifier

MODEL_ID = "Qwen/Qwen3-30B-A3B"

# Load model.
model = Qwen3ForCausalLM.from_pretrained(MODEL_ID, dtype="auto")
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = replace_modules_for_calibration(model)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp8 with per-block quantization
#   * quantize the activations to fp8 with dynamic token activations
recipe = QuantizationModifier(
    targets="Linear",
    scheme="FP8_BLOCK",
    ignore=["lm_head"],
)

# Apply quantization.
oneshot(model=model, recipe=recipe)

# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-block"
model.save_pretrained(SAVE_DIR)
processor.save_pretrained(SAVE_DIR)

Evaluation

The model was evaluated on the OpenLLM leaderboard task, using lm-evaluation-harness. vLLM was used for all evaluations.

Evaluation details

Openllm V1

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-30B-A3B-FP8-block",dtype=auto,add_bos_token=True,max_model_len=16384,tensor_parallel_size=2,gpu_memory_utilization=0.9,enable_chunked_prefill=True,trust_remote_code=True \
  --tasks openllm \
  --write_out \
  --batch_size auto \
  --show_config

Openllm V2

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-30B-A3B-FP8-block",dtype=auto,add_bos_token=False,max_model_len=16384,tensor_parallel_size=2,gpu_memory_utilization=0.7,disable_log_stats=True,enable_chunked_prefill=True,trust_remote_code=True \
  --tasks leaderboard \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --write_out \
  --batch_size auto \
  --show_config

Coding Benchmarks

evalplus.evaluate --model "RedHatAI/Qwen3-30B-A3B-FP8-block" \
                  --dataset "humaneval" \
                  --backend vllm \
                  --tp 2 \
                  --greedy

evalplus.evaluate --model "RedHatAI/Qwen3-30B-A3B-FP8-block" \
                --dataset "mbpp" \
                --backend vllm \
                --tp 2 \
                --greedy

Accuracy

Category	Metric	Qwen/Qwen3-30B-A3B	RedHatAI/Qwen3-30B-A3B-FP8-block	Recovery (%)
OpenLLM V1	ARC-Challenge (Acc-Norm, 25-shot)	69.28	69.88	100.86
GSM8K (Strict-Match, 5-shot)	89.99	89.16	99.07
HellaSwag (Acc-Norm, 10-shot)	77.64	77.41	99.71
MMLU (Acc, 5-shot)	79.50	79.37	99.84
TruthfulQA (MC2, 0-shot)	53.20	53.93	101.38
Winogrande (Acc, 5-shot)	72.30	72.69	100.55
Average Score	73.65	73.74	100.12
OpenLLM V2	IFEval (Inst Level Strict Acc, 0-shot)	48.68	47.84	98.28
BBH (Acc-Norm, 3-shot)	32.46	32.06	98.77
Math-Hard (Exact-Match, 4-shot)	18.81	18.96	100.80
GPQA (Acc-Norm, 0-shot)	24.16	24.75	102.43
MUSR (Acc-Norm, 0-shot)	38.62	40.48	104.79
MMLU-Pro (Acc, 5-shot)	23.15	25.66	110.80
Average Score	30.98	31.62	102.07
Coding	HumanEval pass@1	93.30	93.90	100.64
HumanEval+ pass@1	87.80	88.40	100.68
MBPP pass@1	86.00	85.20	99.06
MBPP+ pass@1	73.00	73.30	100.41