BEIJING (TechGenez) – Chinese artificial intelligence startup DeepSeek released the official API version of its V4 Pro model on Wednesday, delivering a significant leap in AI agent performance.
The new version of the model is now fully compatible with the Responses API and integrates with Codex, opening new possibilities for developers to build sophisticated agent-based applications and complex coding tasks.
DeepSeek priced the V4 Pro at 3 yuan ($0.42) per million input tokens and 6 yuan per million output tokens. Cached inputs are available at just 0.025 yuan per million tokens, maintaining the company’s strategy of offering highly competitive pricing for advanced AI models.
Strong Performance in Agent Benchmarks
The launch highlights DeepSeek’s growing focus on making AI agents that can work independently, write and execute code, and complete multi-step tasks. Benchmark results show the V4 Pro closing the gap with leading models:
- On Terminal-Bench, a test for AI agents in terminal environments, V4 Pro scored 87.9, narrowly behind Claude Fable 5’s 88.
- The model outperformed Fable 5 on CyberGym, a cybersecurity-focused benchmark, and AutomationBench, which evaluates agents on complex automation tasks.
These results position DeepSeek as a serious competitor not only in raw model intelligence but also in the practical ability of AI systems to act autonomously.
Pricing and Accessibility
The new API version is available now to developers worldwide. DeepSeek continues its approach of keeping advanced AI accessible and affordable, which has helped the company gain rapid adoption in coding, research, and agent-building applications.
Strategic Context
DeepSeek’s V4 Pro upgrade comes at a time when the global AI race is intensifying. Chinese developers are increasingly challenging established Western models on both capability and cost, while also pushing AI agents that can operate independently.
The launch follows other recent moves by DeepSeek to expand its API offerings and integrate the model into more developer tools.
Outlook
Analysts expect the V4 Pro API to attract significant attention from startups and enterprises looking for cost-effective AI agents. As the model continues to improve in real-world task completion, it could become a go-to choice for building practical AI automation systems.
Conclusion
DeepSeek’s release of the V4 Pro model marks another important step in the company’s push to make powerful AI agents more accessible and capable. With strong benchmark results and attractive pricing, the upgrade helps Chinese AI developers compete head-to-head with global leaders in the rapidly evolving field of agent-based artificial intelligence.
