DeepSeek-V4-Pro Weights Released, Focusing on Efficient Million-Token Context

multimodalart · x · 2026-08-13

DeepSeek has released the weights for DeepSeek-V4-Pro-0813 on Hugging Face. According to the model card, it focuses on "highly efficient million-token context intelligence" for text generation and supports 8-bit and fp8 precision. The model is available under the MIT license.

Related event: DeepSeek-V4-Pro Officially Released with Enhanced Agent Capabilities and Peak Pricing(48 posts)→

Original post →

More from Models

Models channel →