DeepSeek-V4.1-Flash puts 1M tokens on MIT weights

DeepSeek released V4.1-Flash, a multimodal mixture-of-experts model with 552 billion backbone parameters, 8 billion active input parameters, 16 billion active output parameters, and a context window of up to 1 million tokens. Its API announcement says the model is available with lower peak and off-peak rates; the MIT-licensed model card publishes local-inference and benchmark material. Builders should treat it as a routing candidate rather than an automatic replacement: long contexts and sparse activation still need measured latency, cache behavior, and task quality beside DeepSeek’s time-of-day API pricing.