DeepSeek V4 launches mid-July with 1M-token context and tiered API pricing

DeepSeek has set a mid-July launch date for the official release of DeepSeek V4, building on the preview version with enhanced performance and new features. The entire model lineup will now come with a standard 1-million-token context window, significantly expanding the ability to process long documents and complex multi-step tasks. According to the company, V4 delivers stronger performance in agent-based task execution, mathematical reasoning, and code generation, areas critical for enterprise automation and developer workflows.

Alongside the release, DeepSeek introduced a novel pricing structure for its API. For the first time, the company will implement peak and off-peak pricing. Peak hours are defined as 9:00 AM to 12:00 PM and 2:00 PM to 6:00 PM daily, during which API usage will be charged at double the off-peak rate. This shift aims to manage server load and incentivize usage during lower-demand periods. The pricing change reflects growing infrastructure costs and a move toward dynamic resource allocation common among major cloud providers.