DeepSeek V4 Leak: GPT/Claude Programming Crisis During Spring Festival

🧨With the Spring Festival approaching, DeepSeek is about to shock the world once again this year🧨 📄According to foreign media outlet The Information, two sources directly familiar with the plan revealed that DeepSeek will release V4 around mid-February, just before or after the Spring Festival, although the date may be adjusted. 💪DeepSeek-V4 focuses on coding capabilities, and preliminary internal testing shows it has surpassed other existing models such as Anthropic's Claude and OpenAI's GPT series. The two insiders also added that the core breakthroughs of V4 are reflected in two aspects: 1️⃣Achieving a key breakthrough in handling and parsing ultra-long code prompts. 2️⃣Throughout the entire training process, the model's data pattern understanding ability does not show performance degradation, and is significantly improved compared to previous generations. PS: In the training process of AI models, the model is required to repeatedly learn from massive datasets. However, in practice, as the number of training epochs increases, the model's ability to capture data patterns often declines. For developers with a large reserve of AI chips, the conventional way to address this problem is by increasing the number of training epochs to compensate for performance loss. 🧐In actual use, users are likely to find that answers generated by V4 are more logically clear and structurally organized. This indicates that the model possesses stronger deep reasoning abilities and its reliability in handling complex tasks will also be greatly improved. Notably, some netizens observed that the DeepSeek-V3.2 paper mentioned using the large model arena platform (ChatbotArena) for human preference evaluation. Therefore, we may be able to test this model on the large model arena even earlier.