by txn545
Open source · 105k downloads · 24 likes
The Qwen3.5 122B A10B NVFP4 model is a quantized version of Alibaba’s Qwen3.5-122B-A10B, optimized for efficient execution on NVIDIA GPUs. It is an autoregressive language model built on a transformer architecture, capable of processing textual inputs, images, and videos to generate structured text responses. Designed for diverse applications such as conversational agents, retrieval-augmented generation (RAG) systems, or chatbots, it stands out for its ability to handle extremely long contexts—up to 262,000 tokens—while delivering enhanced performance through dedicated hardware optimization. Its Apache 2.0 license facilitates both commercial and non-commercial use, and its compatibility with NVIDIA Blackwell GPUs ensures reduced inference times. This model serves as a ready-to-use solution for developers seeking to deploy high-performance AI applications without requiring complex configurations.
The txn545/Qwen3.5-122B-A10B-NVFP4 model is a quantized version of Alibaba's Qwen3.5-122B-A10B model, an auto-regressive language model that uses an optimized transformer architecture. For more information on the base model, please check here. This model was quantized using the NVIDIA Model Optimizer.
This model is ready for commercial/non-commercial use.
This model is built upon a third-party base model; see the link to the Non-NVIDIA (Qwen3.5-122B-A10B) Model Card for original requirements and use cases.
Global
Developers looking to take off-the-shelf, pre-quantized models for deployment in AI Agent systems, chatbots, RAG systems, and other AI-powered applications.
Huggingface 02/25/2026 via https://huggingface.co/txn545/Qwen3.5-122B-A10B-NVFP4
Architecture Type: Transformers
Network Architecture: Qwen3.5-122B-A10B
Number of Model Parameters: 122B in total and 10B activated
Input Type(s): Text, Image, Video
Input Format(s): String, Red, Green, Blue (RGB), Video (MP4/WebM)
Input Parameters: One-Dimensional (1D), Two-Dimensional (2D), Three-Dimensional (3D)
Other Properties Related to Input: Context length up to 262K
Output Type(s): Text
Output Format: String
Output Parameters: 1D (One-Dimensional): Sequences
Other Properties Related to Output: N/A
This AI model is designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g., GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.
Supported Runtime Engine(s):
Supported Hardware Microarchitecture Compatibility:
Preferred Operating System(s):
The model is quantized with nvidia-modelopt v0.43.0.dev52+g35e60991c
** Link: cnn_dailymail, Nemotron-Post-Training-Dataset-v2
** Data Collection Method by dataset: Automated.
** Labeling method: Automated.
** Properties: The cnn_dailymail dataset is an English-language dataset containing just over 300k unique news articles as written by journalists at CNN and the Daily Mail.
** Data Modality: Undisclosed
** Data Collection Method by dataset: Undisclosed
** Labeling Method by dataset: Undisclosed
** Properties: Undisclosed
** Data Collection Method by dataset: Undisclosed
** Labeling Method by dataset: Undisclosed
** Properties: Undisclosed
** Data Collection Method by dataset: Hybrid: Human, Automated
** Labeling Method by dataset: Hybrid: Human, Automated
** Properties: Evaluated on benchmarks including GPQA, which is a dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry.
Engine: SGLang
Test Hardware: B200
This model was obtained by quantizing the weights and activations of Qwen3.5-122B-A10B to NVFP4 data type, ready for inference with SGLang. Only the weights and activations of the linear operators within transformer blocks in MoE are quantized.
To serve this checkpoint with SGLang, you need to use the latest main branch with this PR and run the sample command below:
python -m sglang.launch_server --model txn545/Qwen3.5-122B-A10B-NVFP4 --quantization modelopt_fp4 --trust-remote-code
The base model was trained on data that contains toxic language and societal biases originally crawled from the internet. Therefore, the model may amplify those biases and return toxic responses, especially when prompted with toxic prompts. The model may generate answers that may be inaccurate, omit key information, or include irrelevant or redundant text producing socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive.
Trustworthy AI is a shared responsibility. When downloaded or used in accordance with the Apache 2.0 license, developers should work to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.