[DeepSeek Open-Sources V4 Model with 1.6 Trillion Parameters] Odaily Planet Daily reports that DeepSeek has released a preview version of its V4 series open-source models under the MIT license. The weights are now available on Hugging Face and ModelScope. The series includes two MoE models: V4-Pro, with a total of approximately 1.6 trillion parameters and 49 billion parameters activated per token, and V4-Flash, with a total of 284 billion parameters and 13 billion parameters activated. Both models support a context length of up to 1 million tokens. According to the official statement, compared to version 3.2, the new models significantly reduce memory usage and computational overhead during long-text inference.