NVIDIA Releases Quantized DeepSeek-V4.1-Flash-1017 NVFP4 Model

The NVIDIA DeepSeek-V4.1-Flash-1017 NVFP4 model is a quantized iteration of DeepSeek AI’s original DeepSeek-V4.1-Flash-1017 model, featuring an optimized transformer architecture with a Causal Encoder-Decoder (CED) and Compressed Sparse Attention 2 (CSA2). This version supports a context window of up to one million tokens and is designed for both commercial and non-commercial use under the MIT license. The model is natively multimodal, accepting text and image inputs, and can be deployed using the vLLM runtime engine with specific MXFP4 Engram support. NVIDIA states that by leveraging their GPU-accelerated systems and CUDA libraries, the model achieves faster training and inference times compared to CPU-only solutions. The model is available globally and is particularly suitable for AI Agent systems, chatbots, and other AI-powered applications.