DeepSeek-V4-Pro GA Release

DeepSeek has launched DeepSeek-V4-Pro, a model designed for production-grade agentic workflows and complex reasoning tasks. This release introduces flexible reasoning effort controls and native integration with the OpenAI Responses API to streamline developer adoption.

Enhanced Agent Capabilities and Reasoning Control

DeepSeek-V4-Pro provides significant performance gains for production agent workflows. To optimize the balance between latency and accuracy, the model introduces a flexible "reasoning effort" setting available for both V4-Pro and V4-Flash.

Users can configure the reasoning effort based on the task complexity:

  • Low: Optimized for simple tasks.
  • High: Designed for daily agent workflows.
  • Max: Reserved for the most complex tasks requiring deep reasoning.

API Integration and Ecosystem Support

DeepSeek-V4-Pro is natively compatible with the OpenAI Responses API, which is optimized for Codex integrations. This allows developers to implement the model via a one-click setup process.

Availability and Access

DeepSeek-V4-Pro is available through the following channels:

  • Web and App: Accessible via "Expert Mode".
  • API: Available using existing model names; setup details are provided in the official API documentation.

Updated API Pricing Model

Effective August 16, 2026, at 16:00 UTC, DeepSeek is updating its API pricing structure. The new system introduces peak and off-peak rates to encourage flexible workload scheduling. Off-peak rates are 50% lower than peak rates.

Sources

Related

  • Dispatch
    DeepSeek-V4 Preview Release

    DeepSeek has released the open-source DeepSeek-V4 preview, introducing two model variants—Pro and Flash—featuring a 1M token context window and state-of-the-art agentic coding capabilities.

  • Dispatch
    DeepSeek V4 Pro 0813 Release: Performance, Pricing, and Benchmarks

    DeepSeek V4 Pro 0813 offers high-tier reasoning and scientific capabilities at a significantly lower cost than competitors, though user feedback indicates mixed results in coding reliability compared to its Flash counterpart.

  • Dispatch
    DeepSeek-V3.2 and DeepSeek-V3.2-Speciale Release

    DeepSeek has released DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, reasoning-first models designed for agents that introduce integrated thinking in tool-use and gold-medal performance in competitive programming and mathematics.

  • Dispatch
    DeepSeek-V4-Flash Update

    DeepSeek has released a public beta update for DeepSeek-V4-Flash, significantly enhancing agent capabilities and benchmark performance through re-post-training while maintaining the original model architecture.