DeepSeek has announced the General Availability (GA) release of DeepSeek-V4-Pro, featuring significant agent upgrades and production gains.

The new V4-Pro and V4-Flash models offer flexible reasoning effort, allowing for low effort on simple tasks, high effort for daily agent workflows, and maximum effort for complex tasks. A key enhancement is native Open AI Responses API support, optimized for Codex with a one-click setup.

DeepSeek-V4-Pro is accessible via app/web, including an "Expert Mode," and through its API. Model names remain the same, with setup details available in the API documentation.

Alongside the V4 lineup, DeepSeek is updating its API pricing with the introduction of peak and off-peak rates. Off-peak rates will be 50% lower than peak rates, encouraging more flexible workload scheduling. This new pricing structure will be effective starting 16:00 UTC, Aug 16, 2026.