Tencent Releases WeMM-Video-2B-GGUF Quantized Model

At a Glance
| Item | Value |
|---|---|
| Repository | tencent/WeMM-Video-2B-GGUF |
| Family guide | WeMM-Video guide (2 articles) |
| Publisher guide | Tencent: models and licenses |
| Published | 2026-11-17 |
| License | apache-2.0 |
| Formats | GGUF |
| Source type | Primary source (the publisher itself) |
Values determined by this site’s code when the information was collected. Dates are JST.
Overview
Tencent has released “tencent/WeMM-Video-2B-GGUF", a GGUF Q4_K_M quantized package for “WeMM-Video-2B", a model designed for integrated audio and visual understanding and inference. This model is a bilingual model supporting Chinese and English, accepting text, image, audio, and video inputs to generate text-format responses. Provided for the WeMM native runtime, it supports processing video frames alone as well as processing combinations of video frames and original audio.
Specifications
- Parameters: 2B (Dense)
- Architecture: WeMMVideoForConditionalGeneration
Strengths and Use Cases
The features and strengths of this model are based on the capabilities of the original “WeMM-Video-2B" model. It is suited for tasks that integrate and understand visual information from images and videos along with audio information to perform inference.
Specifically, it supports the following use cases:
- Integrated Audio-Visual Understanding and Inference: It is capable of describing events occurring on a video screen, transcribing speech from audio tracks, and summarizing or judging situations based on both visual and auditory information.
- Support for Diverse Input Formats: It supports not only videos with audio, but also videos without audio, single audio files, still images, and text-only inputs.
- Bilingual Support: It supports output in Chinese and English.
- Thinking Mode Switching: Direct answers can be obtained by specifying
enable_thinking=Falsewhen applying prompts, while settingenable_thinking=Trueallows the model to output responses containing the reasoning process (Thinking mode).
Please note that because this package is a quantized version, it is lightweight compared to the original model, but it is reported that accuracy may be somewhat affected.
Hardware Requirements
Estimated requirements (calculated by Local Model Watch) — 3.3B parameters (taken from the base model tencent/WeMM-Video-2B)
| Your VRAM | Quantization | File size | Est. memory needed |
|---|---|---|---|
| 8GB (RTX 4060 / 3060 Ti, etc.) | F32 | 6.1GB | 7.3GB |
Inference engine support (architecture name matched against each project’s own model registry in its source code, checked 2026-11-20): llama.cpp: not registered, vLLM: not registered, MLX (mlx-lm): not registered. “Not registered" means the name is absent from that registry today, not that the model cannot run.
Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. The listed files are distributed in F32 (32-bit). Loading in a lower precision (BF16 or quantized) needs less memory, so treat this as the size of the files as distributed, not a minimum. This site has not measured the lower-precision figures. Compare with other models in our VRAM quick reference. What the quantization names mean: glossary.
Can You Run It Locally?
Runs in Ollama, LM Studio and llama.cpp as-is.
It is distributed in GGUF, so no conversion is needed.
License — apache-2.0 (Commercial use allowed): Permits commercial use, modification and redistribution. Redistribution requires including the license and stating changes; includes a patent grant.
Compiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found" means we have not seen such a build, not that none exists. License summaries are not legal advice — check the publisher’s original terms before relying on them.
Recent Models in the Same Size Class
Models with up to 4B parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.
→ Scroll horizontally to see all columns
| Model | Parameters | Smallest VRAM tier | License | Our article |
|---|---|---|---|---|
| LiquidAI/d1-omni-600M | 587M | 4GB | lfm1.0 | d1-omni-600M Multimodal Decision Model: 4GB+ VRAM, GGUF Builds (2026-11-17) |
| LiquidAI/d1-3B | 3.1B | 4GB | lfm1.0 | d1-3B Vision-Language Model: 4GB+ VRAM, GGUF Builds (2026-11-17) |
| baidu/Unlimited-OCR-2 | 3.3B | 8GB | mit | Unlimited-OCR-2 Vision-Language Model: 8GB+ VRAM (2026-11-17) |
| openbmb/MiniCPM-OCR-1.0-DSpark | 297M | 4GB | apache-2.0 | MiniCPM-OCR-1.0-DSpark OCR Acceleration Draft Model: 4GB+ VRAM (2026-11-17) |
| openbmb/MiniCPM-OCR-Jev-GPTQ | 2.2B | 4GB | apache-2.0 | OpenBMB Releases Quantized Versions of MiniCPM-OCR-Jev (2026-11-17) |
How to Get It
This quantized model is available from the Hugging Face repository tencent/WeMM-Video-2B-GGUF. It is released under the Apache-2.0 license, and no special consent application or agreement to terms of use (Gated) is required for downloading.
The distribution format is GGUF (Q4_K_M), consisting of the following three files split by role. When running, these three files must be used together from the same revision.
language.gguf: Language model, tokenizer, chat templatemmproj.gguf: Vision encoder and projectoraudio.gguf: Audio encoder and projector
To run this package, build the WeMM native runtime and install the huggingface_hub Python package. Running the native/run_hf.py script from the root of the WeMM-Video repository will automatically fetch the required components from the Hub.
An example command for text-only inference is as follows:
export PATH="$PWD/native/install-cpu/bin:$PATH"
MODEL=tencent/WeMM-Video-2B-GGUF
python native/run_hf.py --repo-id "$MODEL" \
--prompt "A 20-minute meeting starts at 09:45. When does it end?" --output text-result.json
When inputting media files, specify the following options according to your purpose:
- Image input:
--image - Audio input:
--audio - Video frames only:
--video(limit frame count with--max-frames) - Video with audio:
--video-audio
An example command to input a video with audio, describe the scene, and summarize the audio is as follows:
python native/run_hf.py --repo-id "$MODEL" --video-audio "examples/media/apollo-11-first-steps.mp4" \
--max-frames 16 --prompt "Describe the scene and summarize the speech." --output video-audio-result.json
Related Articles
- WeMM-Video-2B-NVFP4 Vision-Language Model: 8GB+ VRAM
- Qwen3.8-27B-DSpine-b16 Draft Model for Speculative Decoding: 8GB+ VRAM
- Hy4 Model Specifications, Benchmarks, and Use Cases
- Tencent Releases Hy4 flash: A Cost-Effective Multimodal MoE Model
What to Read Next
- Find models by VRAM (This model runs from the 8GB tier) → Other models that run on a 8GB GPU
- Explore the same model family → WeMM-Video family overview (2 articles)
- Engines that run this model → llama.cpp / Ollama
- Formats this model is available in → GGUF format guide and models
- Learn about the publisher → Tencent: models, licenses and articles
