DeepSeek V4-Flash Sets a New Benchmark for Affordable AI

DeepSeek’s V4-Flash model is the cheapest AI model to run, research firm Artificial Analysis has found. The firm compared AI models by measuring the cost required to complete its Intelligence Index test suite.

Released on July 31, V4-Flash costs about 3 cents per benchmark test compared to the 86 cents charged by Moonshot AI’s Kimi K3. OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Fable cost much more at $1.86 and $3.15, respectively.

What is DeepSeek’s V4-Flash AI Model?

V4-Flash is DeepSeek’s faster, more affordable model. The AI startup developed V4-Flash alongside V4-Pro for users who prioritize efficiency over advanced reasoning. The new AI model is capable of handling everyday tasks at scale at a lower cost.

The primary trade-off is performance, with benchmark tests showing V4-Flash trails higher-end frontier models. On the Intelligence Index, V4-Flash scored 50/100, which is comparable to Gemini 3.6 Flash. However, V4-Flash trailed Meta’s Muse Spark 1.1 and Z.ai’s GLM-5.2, whose score on this benchmark stands at 51. Frontier AI models score significantly higher on the Intelligence Index.

Kimi K3 scored 57, while Claude Opus 5, Claude Fable 5, and GPT-5.6 scored about nine points higher. These results show that lower cost does not always mean the strongest performance. DeepSeek priced V4-Flash at $0.14 per one million input tokens and $0.28 per one million output tokens.

DeepSeek’s Efficiency and Pricing Advantage

DeepSeek is relying on engineering efficiency and competitive pricing to strengthen its position in the AI market. The company has focused on making training and inference more efficient, allowing it to offer lower prices without relying on unsustainable spending.

This strategy could have implications beyond DeepSeek’s API pricing. Analysts argue that aggressive pricing from Chinese AI firms could challenge valuations of OpenAI and Anthropic IPOs. This is because premium pricing would be difficult to justify when competitors offer similar services at a fraction of the cost.

Flash-class models are built for large-scale use cases, including chatbots, coding tools, and automation systems. DeepSeek’s R1 model disrupted the AI industry in early 2025 after it sparked concerns over heavy AI spending by U.S. companies. Since then, rivals such as Moonshot AI, MiniMax, ByteDance, and Alibaba have challenged DeepSeek’s early advantage.

What V4-Flash Means for the AI Race

The emergence of V4-Flash shows that the future of AI competition will be based on performance alone. As businesses adopt AI, affordability and efficiency of AI models could become a major factor. DeepSeek’s approach challenges the industry’s focus on expensive, high-end systems and could push rivals to improve cost efficiency. While frontier AI models will maintain their leadership in handling complex tasks, lightweight models like V4-Flash may become the backbone of everyday AI applications globally.