Chinese AI Models Challenge US Rivals With Lower Prices and Improved Performance
Chinese artificial intelligence (AI) models are rapidly closing the performance gap with US competitors while offering significantly lower prices, according to benchmarking firm Artificial Analysis.
DeepSeek-R1, a reasoning model developed by Chinese AI start-up DeepSeek, scored 60 on the Artificial Analysis Intelligence Index, ranking third globally. It trailed only OpenAI’s o1 and o3-mini, which scored 62 and 66, respectively. The index evaluates AI models based on intelligence and reasoning capabilities.
DeepSeek-R1 also ranked eighth in pricing for developers, while OpenAI’s GPT-4.5 and o1 were the most expensive. The rankings highlight how Chinese AI models are catching up in performance while aggressively lowering prices amid a domestic price war.
Artificial Analysis noted that a year ago, the AI landscape was dominated by US companies. Now, nearly a dozen Chinese firms have models that match or surpass most US-developed AI models.
Alibaba Group Holding is another key player in China’s AI sector. Its latest reasoning model, QwQ-32B, launched earlier this month, ranked fourth in intelligence and tenth in pricing. Both DeepSeek-R1 and QwQ-32B outperformed Western models such as Anthropic’s Claude 3.7 Sonnet, Mistral AI’s Mistral Large 2, and Amazon’s Nova Pro in both intelligence and cost-effectiveness.
DeepSeek-R1 charges USD 2.19 per million tokens of output for users accessing the service via its application programming interface (API). In contrast, OpenAI’s o1 charges USD 60 per million tokens of output, nearly 30 times more than DeepSeek-R1.
On Thursday, OpenAI introduced o1-pro, an advanced version of the o1 model, through its API platform. The new model, which uses more computing power to enhance response quality, is OpenAI’s most expensive to date, charging USD 150 per million tokens of input and USD 600 per million tokens of output.
DeepSeek disrupted the market earlier this year by launching top-tier models that were reportedly trained at a fraction of the cost and computing resources required by Western competitors. The start-up is part of a broader trend among Chinese AI firms engaged in a fierce price war to attract users.
Alibaba’s cloud computing unit recently began offering DeepSeek models on its large language service platform, Bailian. It provides one million free tokens per user for each of its full-strength V3 and R1 models. Additionally, one of its distilled models is priced at just 0.5 yuan (USD 0.07) per million tokens, which Alibaba claims is the lowest price on the market.
DeepSeek-R1 ranked third globally in AI intelligence, behind OpenAI’s o1 and o3-mini
Chinese AI models are closing the performance gap with US rivals while offering lower prices
DeepSeek-R1 charges USD 2.19 per million tokens, significantly less than OpenAI’s o1 at USD 60
Source: SCMP