DeepSeek Returns to the Spotlight with a Strategy that Could Reshape the Global AI Market
DeepSeek is back in the spotlight with a strategy that could redistribute the cards in the global artificial intelligence market. The Chinese startup officially launched its new V4-Flash model on Friday, presented as the least expensive AI model to operate among the main market references, according to analyses by the independent firm Artificial Analysis.
Key Features of V4-Flash
- The average cost of running benchmark tests with V4-Flash is approximately $0.03 per task, a level significantly lower than that of competing models.
- The model is over 100 times cheaper to use than Anthropic's Claude Fable 5 on certain benchmarks, while remaining competitive in terms of performance.
- DeepSeek charges $0.14 per million input tokens and $0.28 per million output tokens, rates that place the company well below the prices practiced by several major players in the sector.
Potential Impact on Businesses
For companies deploying large-scale AI applications - such as conversational assistants, customer support automation, content generation, or software development - this difference can represent substantial savings.
Performance of V4-Flash
Beyond pricing, V4-Flash boasts a level of performance that keeps it in the group of the most performing models on the market. Artificial Analysis attributes a score of 50 to it on its intelligence index, a result comparable to that of some recent Google models and close to other Chinese references of the latest generation.
Market Context
This announcement comes as DeepSeek seeks to regain its momentum after the arrival of new, particularly aggressive Chinese rivals, including Moonshot AI, MiniMax, Z.AI, ByteDance, and Alibaba. According to Reuters, the startup is also preparing for an initial public offering (IPO), although no official schedule has been announced.
