Model Architecture: Qwen3ForCausalLM
- 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/Qwen3-14B.

Model Optimizations

This model was obtained by quantizing the weights and activations of Qwen/Qwen3-14B 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 nm-testing/Qwen3-14B-FP8-block --tensor_parallel_size 1

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 = "nm-testing/Qwen3-14B-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, Qwen3ForCausalLM

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

MODEL_ID = "nm-testing/Qwen3-14B-FP8-block"

# 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="nm-testing/Qwen3-14B-FP8-block",dtype=auto,add_bos_token=True,max_model_len=16384,tensor_parallel_size=1,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="nm-testing/Qwen3-14B-FP8-block",dtype=auto,add_bos_token=False,max_model_len=16384,tensor_parallel_size=1,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 "nm-testing/Qwen3-14B-FP8-block" \
                  --dataset "humaneval" \
                  --backend vllm \
                  --tp 1 \
                  --greedy
evalplus.evaluate --model "nm-testing/Qwen3-14B-FP8-block" \
                --dataset "mbpp" \
                --backend vllm \
                --tp 1 \
                --greedy

Accuracy

Category	Metric	Qwen/Qwen3-14B	nm-testing/Qwen3-14B-FP8-block	Recovery (%)
OpenLLM V1	ARC-Challenge (Acc-Norm, 25-shot)	69.71	69.80	100.12
GSM8K (Strict-Match, 5-shot)	88.40	88.40	100.00
HellaSwag (Acc-Norm, 10-shot)	79.63	79.52	99.86
MMLU (Acc, 5-shot)	78.85	78.73	99.85
TruthfulQA (MC2, 0-shot)	58.58	58.76	100.30
Winogrande (Acc, 5-shot)	73.56	74.27	100.97
Average Score	74.79	74.91	100.16
OpenLLM V2	IFEval (Inst Level Strict Acc, 0-shot)	48.56	48.80	100.49
BBH (Acc-Norm, 3-shot)	31.64	30.76	97.20
Math-Hard (Exact-Match, 4-shot)	19.79	19.56	98.85
GPQA (Acc-Norm, 0-shot)	25.00	24.83	99.33
MUSR (Acc-Norm, 0-shot)	39.02	39.15	100.34
MMLU-Pro (Acc, 5-shot)	24.75	23.42	94.63
Average Score	31.46	31.09	98.82
Coding	HumanEval pass@1	87.80	88.40	100.68
HumanEval+ pass@1	84.80	85.40	100.71
MBPP pass@1	87.00	86.50	99.43
MBPP+ pass@1	75.40	74.10	98.28

RedHatAI/Qwen3-14B-FP8-block

Qwen3-14B-FP8-block

Model Overview