DeepSeek V4 Pro
🧠 AI Modeldeepseek
DeepSeek's MoE model with 1.6T parameters and 1M context for advanced reasoning.
DeepSeek V4 Pro is a large-scale Mixture-of-Experts (MoE) model developed by DeepSeek. It has 1.6 trillion total parameters, with only 49 billion activated per token, making it highly efficient for its size. The model supports a context window of 1,048,576 tokens, enabling it to handle extremely long documents and conversations. It is designed for advanced reasoning, coding, and other complex tasks, as evidenced by its strong ELO scores on benchmarks like fullstack (948, rank #28) and godotgamedev (1098, rank #20). The model offers features such as frequency_penalty, logit_bias, logprobs, and reasoning, making it versatile for various text-based applications. It is available via the OpenRouter platform with text-only input and output modalities.
💡Highlights
- ├─1.6T MoE, 49B activated per token
- ├─1M token context window
- └─Top ELO scores on fullstack and coding benchmarks
🎯For
- ├─AI researchers
- ├─developers
- └─enterprises