DeepSeek V4 Pro (API) AI features and pricing (October 2026)

DeepSeek V4 Pro is the larger model behind DeepSeek's API, a reasoning model with one million tokens of context and thinking mode on by default, offered at comparatively low token prices.

The API model deepseek-v4-pro currently runs DeepSeek-V4-Pro-0813 with a context of one million tokens and up to 384,000 output tokens. It supports JSON output and tool calls, and the API offers endpoints in the format of the Responses API and the Anthropic API, which eases switching from other providers. The V4 series is published with open weights under the MIT license on Hugging Face. Pricing falls by half during off-peak hours.

• Reasons through complex questions cheaplyThinking is on by default, so the model works through problems before answering. Teams test reasoning-heavy workloads at a fraction of typical frontier prices.
• Processes very long inputsOne million tokens of context fit large document sets. Analysts query a whole archive in one request.
• Drops into existing codeEndpoints in Responses API and Anthropic API format let developers switch with small code changes. Teams can compare providers quickly.

• Off-peak pricing at half the peak rate, with peak hours defined in UTC.
• The same model family is available as open weights under the MIT license for self-hosting.

The operator is Hangzhou DeepSeek Artificial Intelligence Co., Ltd., registered in China.

DeepSeek states that it collects, processes and stores personal data in the People's Republic of China. For personal or confidential data, consider self-hosting the open-weight V4 models instead of the API. Check export and procurement rules that apply to your organization.

No reviews yet. Be the first to share how your team uses it.

Every review shows its verification level. We read each one before it goes live. How we check reviews