Chinese AI company DeepSeek has officially launched the production version of its flagship DeepSeek-V4-Pro model, signaling a renewed push to strengthen its position in the rapidly evolving artificial intelligence market.
The new model, known as V4-Pro-0813, introduces significant improvements in reasoning, agent workflows, and real-world task execution. DeepSeek states the release will provide improved performance across coding, automation, and complex problem solving. It also preserves competitive pricing.
Among the key upgrades is a flexible reasoning system that allows users to choose between different levels of computational effort. A low-effort mode is optimized for simple requests. Higher reasoning modes suit advanced workflows and tasks needing deeper analysis.
DeepSeek has also added native compatibility with the OpenAI-style Responses API, enabling easier integration with development tools and agent frameworks. The company says the model supports streamlined deployment for coding assistants and automation platforms with minimal setup requirements.
DeepSeek’s flagship model is available through the API platform, mobile app, and web interface. Users can access an enhanced Expert Mode for more advanced use cases. With the launch, the company announced updated pricing for V4 Pro and the lightweight V4 Flash model. It includes peak and off-peak rates to balance demand.
The release follows months of testing and optimization. During the preview phase, DeepSeek’s lower-cost V4 Flash model outperformed an earlier V4 Pro preview in several independent coding and agent benchmarks. The results guided subsequent development decisions. The company responded by accelerating development efforts and refining the Pro model ahead of its official launch.
DeepSeek is positioning V4 Pro as a high-performance yet cost-efficient alternative in the growing AI market. Industry reports indicate pricing of about $0.435 per million input tokens. The cost for output tokens is $0.87 per million, placing it among the most competitively priced advanced models.
The launch comes as DeepSeek expands aggressively across multiple fronts, including infrastructure, talent acquisition, and fundraising. Reports suggest the company is exploring a new funding round that could value the business at nearly $74 billion, following a major capital raise earlier this year.
Competition in China’s AI sector has intensified, with companies such as Moonshot AI, ByteDance, Alibaba, and Baidu racing to develop increasingly capable large language models. DeepSeek’s latest release is widely viewed as an effort to regain momentum and challenge both domestic competitors and leading international AI providers.
Industry analysts will now be watching closely to see how V4 Pro performs against other frontier AI models, particularly in agentic workflows, software development, and long-context reasoning tasks.
