Qwen3-VL-235B-A22B-Instruct-FP8-dynamic
Model Overview
- Model Architecture: Qwen3VLMoeForConditionalGeneration
- Input: Text/Image/Video
- Output: Text
- Model Optimizations:
- Weight quantization: FP8
- Activation quantization: FP8
- Release Date: 09/28/202510
- Version: 1.0
- Model Developers:: Red Hat
Quantized version of Qwen/Qwen3-VL-235B-A22B-Instruct.
Model Optimizations
This model was obtained by quantizing the weights and activations of Qwen/Qwen3-VL-235B-A22B-Instruct 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-VL-235B-A22B-Instruct-FP8-dynamic --tensor_parallel_size 4
- Send requests to the server:
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-VL-235B-A22B-Instruct-FP8-dynamic"
messages = [
{
"role": "user",
"content": [
{
"type": "image_url",
"image_url": {"url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"},
},
{"type": "text", "text": "Describe this image."},
],
}
]
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.
from transformers import AutoProcessor, Qwen3VLMoeForConditionalGeneration
from llmcompressor import oneshot
from llmcompressor.modeling import replace_modules_for_calibration
from llmcompressor.modifiers.quantization import QuantizationModifier
MODEL_ID = "Qwen/Qwen3-VL-235B-A22B-Instruct"
# Load model.
model = Qwen3VLMoeForConditionalGeneration.from_pretrained(MODEL_ID, torch_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-channel quantization
# * quantize the activations to fp8 with dynamic token activations
recipe = QuantizationModifier(
targets="Linear",
scheme="FP8_DYNAMIC",
ignore=[
"re:.*lm_head",
"re:visual.*",
"re:model.visual.*",
"re:.*mlp.gate$",
],
)
# Apply quantization.
oneshot(model=model, recipe=recipe)
# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-dynamic"
model.save_pretrained(SAVE_DIR)
processor.save_pretrained(SAVE_DIR)
Evaluation
The model was evaluated on the OpenLLMv1 leaderboard task, using lm-evaluation-harness, on reasoning tasks using lighteval and on vision tasks using lmms-eval. vLLM was used for all evaluations.
lm-evaluation-harness
lm_eval \
--model vllm \
--model_args pretrained="RedHatAI/Qwen3-VL-235B-A22B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=4,gpu_memory_utilization=0.8,enable_chunked_prefill=True \
--tasks openllm \
--write_out \
--batch_size auto \
--output_path output_dir \
--show_config
lighteval
lighteval_model_arguments.yaml
model_parameters:
model_name: RedHatAI/Qwen3-VL-235B-A22B-Instruct-FP8-dynamic
dtype: auto
gpu_memory_utilization: 0.9
generation_parameters:
temperature: 0.6
min_p: 0.0
top_p: 0.95
top_k: 20
max_new_tokens: 32768
lighteval vllm \
--model_args lighteval_model_arguments.yaml \
--tasks lighteval|aime25|0 \
lmms-eval
python3 -m lmms_eval \
--model vllm \
--model_args model=RedHatAI/Qwen3-VL-235B-A22B-Instruct-FP8-dynamic,tensor_parallel_size=4,max_model_len=8192,gpu_memory_utilization=0.9 \
--tasks mmmu_val, chartqa\
--batch_size 1