DeepSeek V4.1 Flash 1017: Sparse MoE with 1M-Token Context and 8B/16B Activation
Summary
DeepSeek V4.1 Flash 1017 is a sparse mixture-of-experts model released by DeepSeek with a 1,048,576-token context window and 8B input / 16B output activations from a 552B backbone. The article highlights pricing, performance metrics, and long-horizon task suitability, positioning it as a cost-efficient tier in the V4.1 family.