Microsoft shows how DeepSeek V4 Flash with 284 billion parameters can be executed locally under Windows. Due to extreme quantization, the storage requirement drops to around 60 GB.
This story is only covered by news sources that have yet to be evaluated by the independent media monitoring agencies we use to assess the quality and reliability of news outlets on our platform. Learn more here.
Microsoft shows how DeepSeek V4 Flash with 284 billion parameters can be executed locally under Windows. Due to extreme quantization, the storage requirement drops to around 60 GB.
Microsoft is preparing the local execution on Windows PCs of AI models from 70 to 284 billion parameters: its 1.1 Flash IAM Code, a new Nvidia Nemotron and DeepSeek V4 Flash. Compressed in 1.6 bits, the latter holds in 60 GB of memory. Above all, Microsoft is focusing on hybrid intelligence, capable of using cloud models as well as local models.