Qwen3-VL-235B-A22B-Instruct-FP8-block

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

1. Initialize vLLM server:

vllm serve RedHatAI/Qwen3-VL-235B-A22B-Instruct-FP8-block --tensor_parallel_size 4

2. 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-block"

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.

Creation details

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-block quantization
#   * quantize the activations to fp8 with dynamic token activations
recipe = QuantizationModifier(
    targets="Linear",
    scheme="FP8_BLOCK",
    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-block"
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.

Evaluation details

lm-evaluation-harness

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Qwen3-VL-235B-A22B-Instruct-FP8-block",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-block
  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

Accuracy

Category	Metric	Qwen/Qwen3-VL-235B-A22B-Instruct	RedHatAI/Qwen3-VL-235B-A22B-Instruct-FP8-block	Recovery (%)
OpenLLM V1	ARC-Challenge (Acc-Norm, 25-shot)	76.54	76.19	99.6
	GSM8K (Strict-Match, 5-shot)	90.30	90.07	99.8
	HellaSwag (Acc-Norm, 10-shot)	87.81	87.65	99.8
	MMLU (Acc, 5-shot)	87.11	87.21	100.1
	TruthfulQA (MC2, 0-shot)	63.19	63.41	100.4
	Winogrande (Acc, 5-shot)	81.61	82.08	100.6
	Average Score	81.09	81.10	100.0
Reasoning (generation)	AIME 2025	70.00	76.67	109.5
	Multi-modal	ChartQA (relaxed_overall)	90.12	89.96	99.8
		MMMU (val)	63.67	63.67	100.0
		OpenLLM V2	IFEval (Inst Level Strict Acc, 0-shot)	91.01	90.29	99.21
			BBH (Acc-Norm, 3-shot)	73.72	73.95	100.31
			Math-Hard (Exact-Match, 4-shot)	61.71	20.69	33.54
GPQA (Acc-Norm, 0-shot)			32.13	32.89	102.35
MUSR (Acc-Norm, 0-shot)	42.06		41.80	99.37
MMLU-Pro (Acc, 5-shot)	65.82		65.65	99.73
Average Score	61.07		54.21	88.77