DeepSeek's New AI Model Undercuts Rivals on Cost by Over 100-Fold, Report Finds

Chinese AI startup DeepSeek has released a new artificial intelligence model that is dramatically cheaper to operate than leading global alternatives, a move set to intensify the price war among developers competing for enterprise clients. According to research firm Artificial Analysis, DeepSeek's V4-Flash is by far the least expensive of well-known models to run on benchmark tests.

A New Benchmark for Low Cost

The V4-Flash model, officially released on Friday, is priced at $0.14 per million input tokens and $0.28 per million output tokens, as reported by Artificial Analysis. A token is the basic unit of data processed by AI models. However, a more direct comparison of operational cost shows a starker difference.

Artificial Analysis estimated the average cost to complete a standardized test, which provides a more realistic measure of value by accounting for a model's efficiency.

• DeepSeek V4-Flash: 3 cents
• Moonshot AI Kimi K3: 86 cents
• OpenAI GPT-5.6 Sol: $1.86
• Anthropic Claude Fable 5: $3.15

This pricing makes DeepSeek's model more than 100 times cheaper to run than Anthropic's Claude Fable 5 on this benchmark, highlighting a significant cost advantage for businesses looking to deploy AI at scale.

Performance and Competitive Landscape

Ad

The aggressive pricing comes with a trade-off in performance, positioning the model in the middle tier of capability. Artificial Analysis said the V4-Flash scored 50 out of 100 on its Intelligence Index, which aggregates results from benchmarks measuring coding, reasoning, and other tasks.

This score places it on par with Google's Gemini 3.6 Flash and just one point behind models from Meta and Chinese rival Z.AI. However, it lags behind higher-cost models, with Moonshot’s Kimi K3 scoring a 57, and top-tier offerings from Anthropic and OpenAI scoring nine or more points higher.

Market Implications

DeepSeek, which sources cited by Reuters say is preparing for a potential IPO, is known for its low-cost AI strategy. Its R1 model caused a selloff in global technology stocks in early 2025 by raising questions about the high cost of AI development at major U.S. firms. The launch of V4-Flash appears to be an effort to regain momentum in a crowded field that includes domestic rivals like Moonshot and tech giants such as Alibaba and ByteDance.

This focus on cost efficiency is aimed squarely at enterprise customers seeking affordable ways to integrate AI. DeepSeek has also indicated it is preparing a more powerful version of its model, the V4-Pro, though no release date has been announced.