Enterprise AI costs hit 2026 low driven by price wars, Chinese open-source models: research

2-MIN READ2-MIN Listen

DeepSeek’s new V4-Flash-0731 model costs US$0.03 per task. Photo: Shutterstock Images

Xinmei ShenPublished: 9:30pm, 10 Aug 2026

The cost for businesses to run AI models has fallen to a yearly low, according to research by investment bank Jefferies, driven by a heated global price war and a surge in adoption of low-cost Chinese open-source tools, such as those from DeepSeek.

Average inference prices – measured per million tokens, or chunks of data handled by a model – ranged between US$1.16 and US$1.18 from August 6 to 8. That marked the lowest level recorded this year, Jefferies said on Monday, citing data from US research firm Silicon Data. Average unit costs have plummeted from US$2.04 on May 31 and US$1.45 in late July.

Silicon Data’s index tracks pricing across business application programming interface (API) providers and open-weight inference platforms used by software developers.

The price decline coincided with an “increasing emphasis on cost efficiencies” across both the US and Chinese tech ecosystems, Jefferies analysts led by Thomas Chong noted.

OpenAI escalated the price war last month by slashing rates for its latest GPT-5.6 model series by up to 80 per cent. Similarly, Anthropic’s Claude Opus 5 delivered performance comparable to its flagship Fable 5 model at half the price, Jefferies said in a report last month.