Anthropic Fable 5 training cost is three times that of open source models, facing challenges in the future. In a recent analysis, Wojciech Gryc discussed the market competition between Anthropic’s Fable 5 and two new Chinese models

Anthropic Fable 5 training cost is three times that of open source models, facing challenges in the future. In a recent analysis, Wojciech Gryc discussed the market competition between Anthropic’s Fable 5 and two new Chinese models. Moonshot’s Kimi K3, launched on July 16, and Alibaba’s Qwen 3.8, released on July 19, both achieve similar performance to the Fable 5, and both models will open source their weights. Gryc pointed out that the main cost of training cutting-edge AI models comes from researchers and computing power, but after the model training is completed, the cost of inference is the real overhead. Every time a user calls an API, a lot of GPU resources and power are consumed. He believes that how the cost of reasoning is paid will directly affect the company's business model. He divided AI companies into three categories: the first category is companies that rent data centers and pay electricity bills, such as Anthropic, Moonshot, and GLM 5.2; the second category is companies that build their own data centers, such as Meta and Alibaba; and the third category is companies that have their own power generation capabilities and self-built data centers, such as SpaceX. Gryc emphasized that companies with infrastructure that are not optimal in terms of model performance can still make money by renting out servers. Furthermore, Gryc noted that Anthropic’s Fable 5 is almost three times more expensive than OpenAI and open source models in inference cost for a single task. Whether such high costs can be accepted by users in a highly competitive market is a question worthy of attention. He believes that the current benchmark testing in the market has become saturated, and users’ perceived differences in models are far lower than the cost differences, which may cause Anthropic to face the risk of being abandoned by the market. In contrast, OpenAI shows stronger market competitiveness with its comprehensive product line and data center layout. Gryc predicts that the current wave of competition will continue to develop, and multiple laboratories may continue to surpass traditional U.S. technology giants in the future. Highlights: 🌟 The inference cost of Fable 5 is three times that of OpenAI and open source models, and it faces market pressure. 🔍 Gryc divides AI companies into three categories, pointing out the importance of infrastructure. 📈 Competition continues to intensify, and many Chinese laboratories are quickly catching up or even surpassing American companies. via AI News (author: AI Base)