Qwen3.8-27B

Description

Qwen3.8-27B is a 27-billion-parameter dense vision-language model for text, image, and video understanding. It supports coding, professional-work and research workflows, multimodal reasoning, configurable thinking depth, long-context processing, tool calling, and multi-step agentic tasks.

This model is ready for commercial or non-commercial use.

Third-Party Community Consideration

This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party's requirements for this application and use case; see link to Non-NVIDIA Qwen3.8-27B Model Card from Qwen.

License and Terms of Use:

GOVERNING DOWNLOAD TERMS: Use of this model is governed by the NVIDIA Open Model Agreement.

ADDITIONAL INFORMATION: Apache 2.0.

Deployment Geography:

Global

Use Case:

Developers can use Qwen3.8-27B to build multimodal assistants, coding agents, document and visual-understanding applications, and long-context reasoning workflows that accept text, image, or video inputs and produce text.

Release Date:

Hugging Face: 08/14/2026 via Qwen3.8-27B on Hugging Face

Reference(s):

• Qwen3.8-27B on Hugging Face
• Qwen3.8 announcement
• Qwen3.8 SGLang cookbook
• Qwen3.8 vLLM recipe

Model Architecture:

Architecture Type: Transformer
Network Architecture: Qwen3.8-27B (Qwen3_5ForConditionalGeneration)
Number of model parameters: 27B

Qwen3.8-27B has 64 layers, a hidden dimension of 5,120, and a hybrid layout with 16 repeated groups. Each group contains three Gated DeltaNet and feed-forward units followed by one Gated Attention and feed-forward unit. Gated DeltaNet uses 48 value heads and 16 query/key heads with head dimension 128. Gated Attention uses 24 query heads and four key/value heads with head dimension 256 and rotary-position-embedding dimension 64. The feed-forward intermediate dimension is 17,408, and the model uses multi-token prediction.

Input:

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: The model supports 262,144 tokens of native context and extension up to 1,000,000 tokens. Thinking is enabled by default and can be disabled. Reasoning effort can be configured as xhigh, medium, or low, and thinking content can be preserved across turns.

Output:

Output Type(s): Text
Output Format: String
Output Parameters: One-Dimensional (1D): Sequences
Other Properties Related to Output: The model produces natural-language and code output. Applications can configure sampling and output-length parameters appropriate to their use case.

Our AI models are 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.

Software Integration:

Runtime Engine(s): SGLang
Supported Hardware Microarchitecture Compatibility:

• NVIDIA Blackwell: NVIDIA Blackwell (B200 Tensor Core GPU), NVIDIA Blackwell (RTX PRO 6000D)
• NVIDIA Hopper: NVIDIA Hopper (H200 Tensor Core GPU), NVIDIA Hopper (H20 Tensor Core GPU)

Supported Operating System(s): Linux

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

This AI model can be embedded as an Application Programming Interface (API) call into the software environment described above.

Model Version(s):

• Qwen3.8-27B BF16, source revision 1d4bf0f
• Qwen3.8-27B FP8, source revision 017b9c7

Training, Testing, and Evaluation Datasets:

Training Dataset:

Data Modality: Text, Image, Video
Text Training Data Size: Undisclosed
Image Training Data Size: Undisclosed
Video Training Data Size: Undisclosed
Data Collection Method by dataset: Undisclosed
Labeling Method by dataset: Undisclosed
Properties: The model underwent pre-training and post-training. Training-corpus composition and dataset sizes are Undisclosed.

Testing Dataset:

Data Collection Method by dataset: Undisclosed
Labeling Method by dataset: Undisclosed
Properties: Undisclosed

Evaluation Dataset:

Evaluation Benchmark Score: The Qwen3.8-27B model card provides detailed text and vision-language benchmark results; see the linked third-party model card for the complete score tables and methodology.
Data Collection Method by dataset: [Hybrid: Automated, Manually-Collected]
Labeling Method by dataset: [Hybrid: Automated, Manually-Labeled]
Properties: The evaluation suite covers coding-agent tasks, general reasoning, mathematics, knowledge, image understanding, document understanding, and video understanding. NVIDIA has not independently reproduced the linked results.

Inference:

Acceleration Engine: SGLang
Test Hardware:

• NVIDIA B200 Tensor Core GPU
• NVIDIA H200 Tensor Core GPU
• NVIDIA H20 Tensor Core GPU
• NVIDIA RTX PRO 6000D

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. Developers should work with their internal developer team to ensure these software components meet requirements for the relevant industry and use case and address unforeseen product misuse.

Please make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included.

Users are responsible for model inputs and outputs. Users are responsible for ensuring safe integration of this model, including implementing guardrails as well as other safety mechanisms, prior to deployment.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

Get Help

Getting started with the NIM

Deploying and integrating the NIM is straightforward through industry-standard APIs. Visit the NVIDIA NIM for Vision Language Models documentation for deployment and release guidance.

NVIDIA Developer Community Forum

Get access to community knowledge base articles and support cases through the NVIDIA Developer Forums.