DeepSeek-V4-Pro GA officially announced!
Yesterday's leaked benchmarks were real:
The model now supports adjustable reasoning effort: low for simple tasks, high for everyday agent work, and max for complex problems.
V4-Flash gets the same control.
DeepSeek also added native OpenAI Responses API support and one-click Codex setup, while keeping the existing API model names.
The company claims ""trong production gains" from the agent upgrades, though it has not shared detailed independent benchmarks in the announcement.
V4-Pro is available now via API and in Expert Mode on the DeepSeek app and web.
Pricing is very nice tho
We're launching DeepSeek-V4-Pro today! 🚀 🔷 Major Agent upgrades with strong production gains! 🔷 Flexible reasoning effort for V4-Pro & V4-Flash: low for simp...