WeMM-Video-9B · FP8 Block128

Model Collection Technical Report GitHub Hugging Face Paper

Overview

This repository contains the FP8 Block128 Safetensors variant of WeMM-Video-9B. WeMM-Video is a Chinese–English bilingual family for joint audio-visual understanding and reasoning. It accepts text, images, audio, and video and generates text responses.

  • Weights · 13.10 GiB of weight files
  • Runtime · patched vLLM 0.18.0 on a GPU with FP8 support

Capabilities

WeMM-Video can combine video frames with the original soundtrack to describe events, transcribe speech, and reason about what is seen and heard. It also accepts silent video, standalone audio, images, and text, with responses in Chinese or English.

Examples of WeMM-Video understanding and reasoning across audio-visual, audio, image, and text inputs.

Architecture and training

Vision and audio encoders turn media into tokens that the language model processes alongside text. Video frames carry timestamps. The family was pretrained on approximately 1.2 trillion multimodal tokens, then post-trained with multi-teacher OPD, anti-repetition MrDPO, and specialist reinforcement learning.

WeMM-Video architecture with separate vision and audio encoders feeding media tokens to the language model.

See the project overview for family details.

Quantization

Selected language linear weights use E4M3FN FP8 in 128 × 128 blocks with dynamic per-token activation scaling in groups of 128. The media encoders and protected language tensors retain their source precision. See the quantization manifest for tensor scope and export checks.

Direct loading with Transformers

Use the Transformers environment. The local release checkpoint passed full weight loading and generation on one NVIDIA L20 with Transformers 5.2.0 and PyTorch 2.10.0+cu129. Generation checks covered text, image, audio, video frames, and video with its soundtrack. The same loader accepts the model's Hub ID. The validation used local checkpoint files.

The loaded model retains FP8 weights in the quantized layers.

import torch
from transformers import AutoModelForImageTextToText, AutoProcessor

model_id = "tencent/WeMM-Video-9B-FP8"
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForImageTextToText.from_pretrained(
    model_id, trust_remote_code=True, dtype=torch.bfloat16,
    device_map="cuda:0",
).eval()

messages = [{"role": "user", "content": [
    {"type": "text", "text": "What is 2 + 2? Reply with the number only."},
]}]
inputs = processor.apply_chat_template(
    messages, tokenize=True, return_dict=True, return_tensors="pt",
    add_generation_prompt=True, enable_thinking=False,
).to(device=model.get_input_embeddings().weight.device, dtype=torch.bfloat16)
with torch.inference_mode():
    output = model.generate(**inputs, do_sample=False, max_new_tokens=32)
print(processor.tokenizer.decode(output[0, inputs.input_ids.shape[1]:], skip_special_tokens=True))

Transformers input types

For a media request, place the media item before the text item in messages[0]["content"] and run the same preprocessing and generation.

Input Example content item
Text {"type": "text", "text": "Describe what you see."}
Image {"type": "image", "image": "image.jpg"}
Audio {"type": "audio", "audio": "speech.wav"}
Video frames {"type": "video", "video": "clip.mp4"}
Video with soundtrack {"type": "video_audio", "video": "clip.mp4"}

See the Transformers examples for media decoding dependencies and sampling options.

Serving

Install the matching WeMM runtime patch. The serving guide lists the dependency profiles and Docker images. Choose one server below.

vLLM

Use patched vLLM 0.18.0 on a GPU with FP8 support. See the vLLM recipe for setup details.

MODEL=tencent/WeMM-Video-9B-FP8
vllm serve "$MODEL" --served-model-name "$MODEL" --trust-remote-code --dtype bfloat16 \
  --host 127.0.0.1 --port 8000 --max-model-len 8192 \
  --no-enable-prefix-caching --mamba-ssm-cache-dtype float32 \
  --limit-mm-per-prompt '{"image":4,"video":2,"audio":2}' \
  --media-io-kwargs '{"video":{"fps":2,"min_frames":4,"max_frames":16,"total_pixels":262144}}'

Input examples for the serving API

With the server running on port 8000, use the bundled request client from the project root in a second terminal. It sends media inline to the OpenAI-compatible chat endpoint.

MODEL=tencent/WeMM-Video-9B-FP8
python examples/chat.py text        --model "$MODEL"
python examples/chat.py image       --model "$MODEL"
python examples/chat.py audio       --model "$MODEL"
python examples/chat.py video       --model "$MODEL"
python examples/chat.py video-audio --model "$MODEL"
Input Request content
Text Text string in messages[].content
Image image_url with image_url.url
Audio audio_url with audio_url.url
Video frames video_url with video_url.use_audio set to false
Video with soundtrack video_url with video_url.use_audio set to true

Set fps and max_frames inside video_url to control sampling for a clip. Set chat_template_kwargs.enable_thinking to true for Thinking mode. See the complete video request.

Chat Modes

The checkpoint template defaults to Instruct. Set enable_thinking=True through the runtime's chat-template options to use Thinking mode. Use the media processor and chat template shipped with this checkpoint.

Validation

Patched vLLM loaded the checkpoint on H20 and completed generation checks for text, image, audio, and video with a soundtrack. The runtime precision audit confirmed the W8A8 path.

Evaluation

See the WeMM-Video-35B-A3B evaluation results for the benchmark comparison in the technical report.

License

Apache License 2.0. See the included license and the collection license and third-party attribution. Third-party components remain subject to their original terms.

Citation

@misc{WeMM-Video,
      title={WeMM-Video: Pushing the Frontiers of Open-Source Omni-Modal Video Understanding}, 
      author={Changli Tang and Jie Yang and Tianyi Wang and Lei Li and Daoze Zhang and Zijie Xin and Zhan Liu and Li Li and Jinchang Xu and Fengyun Rao and Chao Zhang and Jing Lyu},
      year={2026},
      eprint={2611.26409},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2611.26409}, 
}
Downloads last month
-
Safetensors
Model size
10B params
Tensor type
F32
·
BF16
·
F8_E4M3
·
I64
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for tencent/WeMM-Video-9B-FP8

Quantized
(4)
this model

Collection including tencent/WeMM-Video-9B-FP8

Paper for tencent/WeMM-Video-9B-FP8