MUMBAI: The AI price war is getting another shot in the arm. Chinese startup DeepSeek has launched V4.1 Flash, a lower-cost model designed to power AI agents while putting speed, throughput and affordability at the centre of the pitch.
DeepSeek describes V4.1 Flash as the smallest model in its new architecture family, with improvements across capability, inference speed, throughput and scalability. The release comes as developers increasingly look beyond chatbots towards AI agents capable of handling multi-step tasks, where every additional action can translate into another model call and another inference bill.
That makes cost more than a footnote. DeepSeek’s V4 Flash models have already been positioned at substantially lower prices than leading US models, helping Chinese AI developers attract users processing large volumes of workloads. V4.1 Flash takes that strategy further by targeting applications where speed and processing capacity matter alongside model capability.
The company is also clearing the runway for its new model by retiring V4 Pro. According to DeepSeek’s API documentation, from September 14, requests made to V4 Pro will be routed to V4.1 Flash and charged at the Flash model’s rates.
DeepSeek says V4.1 Flash has surpassed V4 Pro on performance, cost, speed and total processing time, effectively turning the new release into a lower-cost workhorse rather than simply another model on the shelf.
The shift comes as pricing becomes an increasingly important battleground in AI. Unlike conventional chatbot use, agentic systems can perform a chain of actions, potentially multiplying the number of times a model is called. A cheaper model can therefore make a significant difference when those calls scale into millions or billions.
Recent OpenRouter data cited by UOL also showed seven of its 10 most-used AI models were Chinese, with low pricing emerging as one factor supporting adoption.
For OpenAI, Anthropic and other frontier AI developers, the pressure is therefore coming from two directions: building models that can do more while making sure they remain affordable enough to be used at scale.
DeepSeek’s latest move underlines how the AI race is evolving. The prize is no longer simply the smartest model in the room; increasingly, it is the model that can get the job done without running up the tab.
