Featured image of post AI Model Pricing Shifts Investor Focus to Applications and Chinese Alternatives

AI Model Pricing Shifts Investor Focus to Applications and Chinese Alternatives

AI Model Pricing Shifts Investor Focus to Applications and Chinese Alternatives

Cost Becomes the New Benchmark

Huatai Securities says competition among large language models is moving beyond leaderboard scores toward a more practical question: how much does similar intelligence cost? OpenAI cut prices for Terra and Luna on July 30, by 20% and 80% respectively, signaling that even leading model providers are using pricing to expand adoption.

Chinese Models Gain Ground

According to the report, DeepSeek V4 Flash 0731 scored 50 on the Artificial Analysis Intelligence Index, only one point below Luna. Its blended price is about $0.06 per million tokens, while average task cost is roughly $0.03. A token is a basic unit of text processed by an AI model, so lower token pricing directly reduces usage costs for developers and enterprises. Huatai highlights Kimi K3 as a strong example of capability among Chinese open-weight models, while DeepSeek V4 Flash sets a new cost floor in the 50-point performance range.

Market Takeaway

As model performance gaps narrow and prices fall, investors may pay more attention to two themes: AI applications that can monetize quickly, and domestic models offering strong value for money.