Qwen3-8B-FP8-dynamic

<h1 style="display: flex; align-items: center; gap: 10px; margin: 0;"> Qwen3-8B-FP8-dynamic <img src="https://www.redhat.com/rhdc/managed-files/Catalog-Validated_model_0.png" alt="Model Icon" width="40" style="margin: 0; padding: 0;" /> </h1> <a href="https://www.redhat.com/en/products/ai/validated-models" target="_blank" style="margin: 0; padding: 0;"> <img src="https://www.redhat.com/rhdc/managed-files/Validated_badge-Dark.png" alt="Validated Badge" width="250" style="margin: 0; padding: 0;" /> </a>

Model Overview

• Model Architecture: Qwen3ForCausalLM
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Activation quantization: FP8
  • Weight quantization: FP8
• 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: 05/02/2025
• Version: 1.0
• Validated on: RHOAI 2.24, RHAIIS 3.2.1
• Model Developers: RedHat (Neural Magic)
• ModelCar Storage URI: oci://registry.redhat.io/rhelai1/modelcar-qwen3-8b-fp8-dynamic:1.5
• Validated on RHOAI 2.24: quay.io/modh/vllm:rhoai-2.24-cuda
• Validated on RHAIIS 3.2.1: http://registry.redhat.io/rhaiis/vllm-cuda-rhel9:3.2.1
• Validated on vLLM: 0.10.0

Model Optimizations

This model was obtained by quantizing activations and weights of Qwen3-8B to FP8 data type. This optimization reduces the number of bits used to represent weights and activations from 16 to 8, reducing GPU memory requirements (by approximately 50%) and increasing matrix-multiply compute throughput (by approximately 2x). Weight quantization also reduces disk size requirements by approximately 50%.

Only weights and activations of the linear operators within transformers blocks are quantized. Weights are quantized with a symmetric static per-channel scheme, whereas activations are quantized with a symmetric dynamic per-token scheme. The llm-compressor library is used for quantization.

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/Qwen3-8B-FP8-dynamic"
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.

<details> <summary>Deploy on <strong>Red Hat AI Inference Server</strong></summary>

podman run --rm -it --device nvidia.com/gpu=all -p 8000:8000 \
 --ipc=host \
--env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
--env "HF_HUB_OFFLINE=0" -v ~/.cache/vllm:/home/vllm/.cache \
--name=vllm \
registry.access.redhat.com/rhaiis/rh-vllm-cuda \
vllm serve \
--tensor-parallel-size 8 \
--max-model-len 32768  \
--enforce-eager --model RedHatAI/Qwen3-8B-FP8-dynamic

</details> <details> <summary>Deploy on <strong>Red Hat Openshift AI</strong></summary>

# Setting up vllm server with ServingRuntime
# Save as: vllm-servingruntime.yaml
apiVersion: serving.kserve.io/v1alpha1
kind: ServingRuntime
metadata:
 name: vllm-cuda-runtime # OPTIONAL CHANGE: set a unique name
 annotations:
   openshift.io/display-name: vLLM NVIDIA GPU ServingRuntime for KServe
   opendatahub.io/recommended-accelerators: '["nvidia.com/gpu"]'
 labels:
   opendatahub.io/dashboard: 'true'
spec:
 annotations:
   prometheus.io/port: '8080'
   prometheus.io/path: '/metrics'
 multiModel: false
 supportedModelFormats:
   - autoSelect: true
     name: vLLM
 containers:
   - name: kserve-container
     image: quay.io/modh/vllm:rhoai-2.24-cuda # CHANGE if needed. If AMD: quay.io/modh/vllm:rhoai-2.24-rocm
     command:
       - python
       - -m
       - vllm.entrypoints.openai.api_server
     args:
       - "--port=8080"
       - "--model=/mnt/models"
       - "--served-model-name={{.Name}}"
     env:
       - name: HF_HOME
         value: /tmp/hf_home
     ports:
       - containerPort: 8080
         protocol: TCP

# Attach model to vllm server. This is an NVIDIA template
# Save as: inferenceservice.yaml
apiVersion: serving.kserve.io/v1beta1
kind: InferenceService
metadata:
  annotations:
    openshift.io/display-name: Qwen3-8B-FP8-dynamic # OPTIONAL CHANGE
    serving.kserve.io/deploymentMode: RawDeployment
  name: Qwen3-8B-FP8-dynamic          # specify model name. This value will be used to invoke the model in the payload
  labels:
    opendatahub.io/dashboard: 'true'
spec:
  predictor:
    maxReplicas: 1
    minReplicas: 1
    model:
      modelFormat:
        name: vLLM
      name: ''
      resources:
        limits:
          cpu: '2'			# this is model specific
          memory: 8Gi		# this is model specific
          nvidia.com/gpu: '1'	# this is accelerator specific
        requests:			# same comment for this block
          cpu: '1'
          memory: 4Gi
          nvidia.com/gpu: '1'
      runtime: vllm-cuda-runtime	# must match the ServingRuntime name above
      storageUri: oci://registry.redhat.io/rhelai1/modelcar-qwen3-8b-fp8-dynamic:1.5
    tolerations:
    - effect: NoSchedule
      key: nvidia.com/gpu
      operator: Exists

# make sure first to be in the project where you want to deploy the model
# oc project <project-name>

# apply both resources to run model

# Apply the ServingRuntime
oc apply -f vllm-servingruntime.yaml

# Apply the InferenceService
oc apply -f qwen-inferenceservice.yaml

# Replace <inference-service-name> and <cluster-ingress-domain> below:
# - Run `oc get inferenceservice` to find your URL if unsure.

# Call the server using curl:
curl https://<inference-service-name>-predictor-default.<domain>/v1/chat/completions
        -H "Content-Type: application/json" \
        -d '{
    "model": "Qwen3-8B-FP8-dynamic",
    "stream": true,
    "stream_options": {
        "include_usage": true
    },
    "max_tokens": 1,
    "messages": [
        {
            "role": "user",
            "content": "How can a bee fly when its wings are so small?"
        }
    ]
}'

See Red Hat Openshift AI documentation for more details.

</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 QuantizationModifier
from llmcompressor.transformers import oneshot
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model
model_stub = "Qwen/Qwen3-8B"
model_name = model_stub.split("/")[-1]

model = AutoModelForCausalLM.from_pretrained(model_stub)

tokenizer = AutoTokenizer.from_pretrained(model_stub)

# Configure the quantization algorithm and scheme
recipe = QuantizationModifier(
    ignore=["lm_head"],
    targets="Linear",
    scheme="FP8_dynamic",
)

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

# Save to disk in compressed-tensors format
save_path = model_name + "-FP8-dynamic"
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/Qwen3-8B-FP8-dynamic",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/Qwen3-8B-FP8-dynamic",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/Qwen3-8B-FP8-dynamic",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/Qwen3-8B-FP8-dynamic
  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>Qwen3-8B </th> <th>Qwen3-8B-FP8-dynamic<br>(this model) </th> <th>Recovery </th> </tr> <tr> <td rowspan="7" ><strong>OpenLLM v1</strong> </td> <td>MMLU (5-shot) </td> <td>71.95 </td> <td>72.30 </td> <td>100.5% </td> </tr> <tr> <td>ARC Challenge (25-shot) </td> <td>61.69 </td> <td>61.60 </td> <td>99.9% </td> </tr> <tr> <td>GSM-8K (5-shot, strict-match) </td> <td>75.97 </td> <td>80.52 </td> <td>106.0% </td> </tr> <tr> <td>Hellaswag (10-shot) </td> <td>56.52 </td> <td>55.95 </td> <td>99.0% </td> </tr> <tr> <td>Winogrande (5-shot) </td> <td>65.98 </td> <td>66.22 </td> <td>100.4% </td> </tr> <tr> <td>TruthfulQA (0-shot, mc2) </td> <td>53.17 </td> <td>52.39 </td> <td>98.5% </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>64.21</strong> </td> <td><strong>64.83</strong> </td> <td><strong>101.0%</strong> </td> </tr> <tr> <td rowspan="7" ><strong>OpenLLM v2</strong> </td> <td>MMLU-Pro (5-shot) </td> <td>34.57 </td> <td>37.82 </td> <td>109.4% </td> </tr> <tr> <td>IFEval (0-shot) </td> <td>84.77 </td> <td>84.56 </td> <td>99.8% </td> </tr> <tr> <td>BBH (3-shot) </td> <td>25.47 </td> <td>27.20 </td> <td>106.8% </td> </tr> <tr> <td>Math-lvl-5 (4-shot) </td> <td>51.05 </td> <td>51.90 </td> <td>101.7% </td> </tr> <tr> <td>GPQA (0-shot) </td> <td>0.00 </td> <td>0.00 </td> <td>--- </td> </tr> <tr> <td>MuSR (0-shot) </td> <td>10.02 </td> <td>10.65 </td> <td>--- </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>34.31</strong> </td> <td><strong>35.35</strong> </td> <td><strong>103.0%</strong> </td> </tr> <tr> <td><strong>Multilingual</strong> </td> <td>MGSM (0-shot) </td> <td>25.97 </td> <td>25.80 </td> <td>99.4% </td> </tr> <tr> <td rowspan="6" ><strong>Reasoning<br>(generation)</strong> </td> <td>AIME 2024 </td> <td>74.58 </td> <td>76.35 </td> <td>102.4% </td> </tr> <tr> <td>AIME 2025 </td> <td>65.21 </td> <td>63.75 </td> <td>97.8% </td> </tr> <tr> <td>GPQA diamond </td> <td>58.59 </td> <td>61.11 </td> <td>104.3% </td> </tr> <tr> <td>Math-lvl-5 </td> <td>97.60 </td> <td>96.60 </td> <td>99.0% </td> </tr> <tr> <td>LiveCodeBench </td> <td>56.27 </td> <td>56.60 </td> <td>100.6% </td> </tr> </table>