Skip to content

DeepSeek V4.1 Flash 1017

Model ID: vllm/release/deepseek-v41-flash-1017 DeepSeek Experimental

AttributeValue
ClassificationGPAI (General-Purpose AI Model)
Risk classLimited Risk (Article 50)
Transparency obligationYes
noris roleDistributor

As a GPAI model, this model falls within the scope of the EU AI Act transparency obligations. noris acts as a distributor and makes the model developer’s information available pursuant to Article 53(2) and (3).

PropertyValue
Model nameDeepSeek V4.1 Flash 1017
ProviderDeepSeek
Model IDvllm/release/deepseek-v41-flash-1017
Parameters552B backbone (MoE), 8B active during prefill, 16B during decode, plus 196B Engram
ArchitectureCausal encoder-decoder MoE with Compressed Sparse Attention 2 (CSA2)
Context length1,048,576 tokens
ModalitiesText + image (multimodal)
CheckpointNVFP4 (routed experts), Engram tables in MXFP4, quantized with NVIDIA Model Optimizer
LicenseMIT
  • AI agents and long-running tasks (Terminal-Bench 2.1: 81.1%)
  • Chatbots and assistants
  • RAG systems
  • Very-long-document analysis (1M context)
  • Image understanding (scientific charts, MMMU-Pro: 78.7%)
Risk FactorCategoryLevel
Model sizeDeepfake potentialMedium
MultimodalImage inputMedium
Open source (MIT)Misuse potentialLow
Experimental tierDeprecation noticeFlagged
Environmental impactModel size (552B backbone)Medium-High

Training data: not fully publicly documented, see the base model card. The NVFP4 checkpoint was calibrated with cnn_dailymail and the Nemotron-Post-Training-Dataset-v2.

Sustainability:

MetricValue
CO₂High (552B backbone)
Optimizationonly 8B (prefill) resp. 16B (decode) active parameters despite the large expert pool

Model card created on November 4, 2026. This model card serves EU AI Act transparency purposes (Regulation (EU) 2024/1689). The information is based on publicly available data from the model provider.