DeepSeek updates its V4 Pro API model and warns of a coming price increase

DeepSeek updated its flagship API model to DeepSeek-V4-Pro-0813 on August 12, 2026, pairing a 1-million-token context window with explicit agent-developer tooling, while warning developers that a broader price increase is coming.

This continues DeepSeek’s V4 line: it follows the DeepSeek-V4-Flash-0731 release on July 31, 2026, and pushes the higher-priced Pro tier further toward agentic and long-context work.

The updated model pairs the 1-million-token context window with tool calling, DeepSeek said in its API documentation. Because the call name is unchanged — deepseek-v4-pro — existing integrations pick up the new version automatically, and the model is also listed on third-party access platforms.

Pricing is $0.435 per million input tokens on a cache miss, dropping to $0.003625 on a cache hit, and $0.87 per million output tokens — roughly a threefold premium over DeepSeek’s Flash tier. The documentation also flags a coming “significant” price increase across DeepSeek’s API services, a notable signal from a provider known for undercutting Western rivals on cost.

DeepSeek has not published independent benchmarks with this update, so the practical gain from the larger context window and agent tooling is not yet verifiable. The price warning is the more consequential line for developers who built on DeepSeek precisely because it was cheap.