DeepSeek-V4-Pro, DeepSeek-V4-F... Note

DeepSeek-V4-Pro, DeepSeek-V4-Flash, and DeepSeek-V4.1-Flash models now available on Amazon SageMaker JumpStart

DeepSeek has made its DeepSeek-V4-Pro, DeepSeek-V4-Flash, and DeepSeek-V4.1-Flash models available on Amazon SageMaker JumpStart. These open-weight Mixture-of-Experts (MoE) language models offer frontier-class performance for reasoning, coding, and agentic tasks. They achieve high efficiency by activating only a fraction of their total parameters per token. DeepSeek-V4-Pro is a 1.6T-parameter MoE model designed for complex reasoning and high-quality text generation, supporting a 1M-token context window. DeepSeek-V4-Flash is a more efficient 284B-parameter MoE model, offering strong performance at lower compute cost with the same 1M-token context window. DeepSeek-V4.1-Flash provides next-generation agentic efficiency with native multimodal support using a novel Causal Encoder-Decoder architecture. This model optimizes cost efficiency for agentic workloads, coding, and vision tasks through its architecture and caching mechanisms. Customers can easily deploy these models via SageMaker JumpStart in the SageMaker console or using the SageMaker Python SDK. This expansion provides AWS customers with a wider range of foundation models for diverse AI applications.