DeepSeek V4 Pro officially launches with enhanced AI agent capabilities
DeepSeek launches V4-Pro with major improvements for AI agent workflows. The model features flexible reasoning effort levels (low, high, max) and native OpenAI Responses API support. Pricing updates introduce peak and off-peak rates with 50% discounts during off-peak hours.
What are the main improvements in DeepSeek V4-Pro?
V4-Pro features significant enhancements for AI agent workflows, including flexible reasoning effort levels (low for simple tasks, high for typical agent workflows, max for complex problems). The model also provides native support for OpenAI Responses API.
How did API pricing change with V4-Pro release?
DeepSeek introduced dynamic pricing with peak and off-peak rates. Off-peak rates are 50% lower than peak rates, enabling more flexible workload scheduling. New pricing went into effect on August 16, 2026 at 16:00 UTC.
Where can I use V4-Pro?
V4-Pro is available in the DeepSeek web interface and app (accessible via Expert Mode) and also through the API with unchanged model names. Users should refer to API documentation for setup details.
- DeepSeek makes its cheapest model more powerful: DeepSeek-V4-Flash — huggingface.co 58 % match
- Ask Gemini Live to Deep Research a complex topic — x.com 55 % match
- Xiaomi unveiled HyperOS 4 with adaptive glass design and AI 2.0 assistant — hyperos.mi.com 55 % match
- DeepSeek V4-Pro3
- DeepSeek5
- OpenAI Responses API
- Codex6