DeepSeek makes V4-Pro 75% price cut permanent, slashes agent costs

DeepSeek, the Chinese AI startup known for cost-competitive large language models, has made the 75% price reduction on its V4-Pro model permanent as of May 29, 2026 — ending the temporary promotional period. This move signals a long-term commitment to affordable pricing for high-volume API users. Additionally, the company introduced a dramatic cut to cache-hit pricing for long-running AI agents, reducing it to just one-tenth of previous rates. Standard API pricing now sits around $0.435 per million input tokens, making V4-Pro one of the most cost-efficient models on the market for enterprise deployments.

With a 1M-token context window, V4-Pro is especially suited for applications that require processing large documents, extended conversation history, or complex multi-step reasoning. The combination of permanent low pricing and drastically reduced agent costs positions DeepSeek as a strong alternative for businesses scaling AI workloads — particularly those dealing with heavy caching patterns, such as customer support bots, code assistants, and research agents. This pricing strategy puts direct pressure on competitors like OpenAI and Anthropic to match DeepSeek's efficiency gains.