The new series model DeepSeek-V3 from Algoway has launched its first version and has been made open source, with the API service also updated, requiring no changes to the interface configuration. The current version of DeepSeek-V3 does not support multimodal input and output. It is reported that DeepSeek-V3 is a self-developed MoE model with 671 billion parameters and 37 billion activated, pretrained on 14.8 trillion tokens. Its performance in various evaluations surpasses that of other open source models such as Qwen2.5-72B and Llama-3.1-405B, and it is on par with the world's leading closed source models like GPT-4o and Claude-3.5-Sonnet.