A recent analysis by a research firm indicates that a version of Chinese startup DeepSeek's primary artificial intelligence model is significantly less expensive to operate on benchmark tests compared to other prominent global models. This model, V4-Flash, is reportedly over 100 times more economical to run than Anthropic's Claude Fable 5. DeepSeek, which market sources suggest is preparing for a potential initial public offering, officially launched its V4-Flash model on Friday. This release represents the company's latest effort to regain market traction by focusing on its established reputation for providing highly affordable AI alternatives.

Cost-Effectiveness Details

DeepSeek's V4-Flash model is priced at $0.14 per million input tokens and $0.28 per million output tokens, as reported by Artificial Analysis. A token serves as a fundamental unit of data for quantifying AI usage. The San Francisco-based firm estimated the average operational cost of V4-Flash at 3 cents per test. In comparison, rival Chinese firm Moonshot AI's Kimi K3 averaged 86 cents, OpenAI's GPT-5.6 Sol cost $1.86, and Claude Fable 5 was $3.15 per test. This comparative analysis offers a more practical assessment of value than mere pricing, as it considers the volume of data an AI model must process and generate to complete a given task. A model with a low advertised price could still incur substantial costs if it requires considerably more steps to produce a response.

Performance Benchmarks

Artificial Analysis reported that DeepSeek's V4-Flash model achieved a score of 50 out of 100 on its Intelligence Index. This index integrates results from nine different benchmarks covering areas such as coding, reasoning, and typical workplace assignments. This score matches Google's Gemini 3.6 Flash. It is also just one point behind Meta's Muse Spark 1.1 and GLM-5.2 from Z.AI, also known as Zhipu. However, Moonshot's Kimi K3 scored higher with 57 points, while Anthropic's Claude Opus 5, Fable 5, and OpenAI's GPT-5.6 models scored nine or more points above V4-Flash.

Competitive Landscape and Future Outlook

DeepSeek previously garnered significant attention in the realm of Chinese AI development but soon faced intense competition from numerous domestic rivals. These competitors include other startups like Moonshot, MiniMax, and Z.AI, alongside technology giants such as ByteDance and Alibaba. All these entities are competing with U.S. technology firms for global adoption, specifically targeting businesses seeking more economical methods to implement AI at scale. DeepSeek is also developing a more advanced iteration of its model, named V4-Pro, though an official release date has not yet been announced. Separately, Alibaba introduced its most extensive and capable artificial intelligence model to date, the Qwen3.8-Max, on Monday. This new offering is comparable in size to a model launched last month by domestic competitor Moonshot AI.