**DeepSeek Undercuts OpenAI and Anthropic on Benchmark Running Costs**
In a significant development in the artificial intelligence sector, Chinese startup DeepSeek has emerged as a leader in cost-effective AI solutions. The company’s latest model, V4-Flash, has been reported as the most affordable AI model to operate among its global competitors, significantly undercutting the costs associated with models from established firms like OpenAI and Anthropic.
According to research conducted by Artificial Analysis, DeepSeek’s V4-Flash model charges a mere $0.14 per million input tokens and $0.28 per million output tokens. This pricing translates to an average operational cost of just 3 cents per test. In contrast, the costs for other leading models are substantially higher: Moonshot AI's Kimi K3 runs at 86 cents per test, OpenAI's GPT-5.6 Sol at $1.86, and Anthropic's Claude Fable 5 at $3.15. These figures highlight DeepSeek's competitive edge in the AI market, particularly for businesses looking for more economical solutions.
The term "token" refers to a unit of data that AI models use to process and generate information, making it a critical metric for evaluating AI usage and costs. The comparison of operational costs provides a clearer understanding of value than pricing alone, as it considers the amount of data each model processes to complete tasks.
DeepSeek's rise to prominence began with its earlier model, R1, which gained considerable attention in early 2025, prompting a selloff in global technology stocks and raising concerns about the expenditures of U.S. companies on AI technologies. The release of the V4-Flash model last week signifies DeepSeek's strategic move to regain market momentum by continuing its tradition of providing ultra-low-cost AI alternatives.
Despite its current success, DeepSeek faces stiff competition from a variety of domestic rivals, including startups like Moonshot, MiniMax, and Z.AI, as well as tech giants such as ByteDance and Alibaba. These companies are all vying for global adoption of their AI technologies, targeting businesses that seek cost-effective ways to implement AI on a large scale.
In terms of performance, DeepSeek's V4-Flash model scored 50 out of 100 on the Artificial Analysis Intelligence Index, which evaluates models based on nine benchmarks that include coding, reasoning, and workplace-style tasks. This score is on par with Google’s Gemini 3.6 Flash and just one point behind Meta’s Muse Spark 1.1 and Z.AI's GLM-5.2. However, it falls short of Moonshot's Kimi K3, which scored 57, and Anthropic’s Claude Opus 5 and Fable 5, as well as OpenAI's GPT-5.6, which all scored significantly higher.
Looking ahead, DeepSeek is reportedly preparing to release a more powerful version of its AI model, dubbed V4-Pro, although a specific release date has yet to be announced. This move indicates DeepSeek's commitment to innovation and its intention to maintain its competitive edge in the rapidly evolving AI landscape.
In a related development, Alibaba has introduced its largest and most capable AI model to date, the Qwen3.8-Max, which is positioned closely in size to a recent offering from Moonshot AI. This unveiling underscores the intensifying competition within the AI sector, as companies race to develop advanced models that can meet the growing demand for AI solutions across various industries.
As the AI market continues to expand, the focus on operational costs and performance metrics will likely play a crucial role in shaping the strategies of companies like DeepSeek, OpenAI, and Anthropic. The ongoing competition among these firms will not only drive innovation but also influence the pricing and accessibility of AI technologies for businesses worldwide.