Chinese artificial intelligence (AI) firm Moonshot AI has halved the price of a new feature on its Kimi large language model (LLM), as start-ups compete with the country’s technology giants to monetise their generative AI products.
The price of the context caching feature on Moonshot’s Kimi has been cut to 5 yuan (US$0.70) per 1 million-tokens per minute from 10 yuan, Moonshot said in a blog post on Wednesday.
Context caching lets LLM developers store for a period of time certain information that might be frequently requested so their model responds faster to similar queries.
The feature could “significantly improve efficiency and lower costs”, Moonshot said when it launched public testing of the feature in July.
Moonshot is the latest generative AI firm in China to cut service prices amid fierce competition in the country, where Big Tech and start-ups are racing to commercialise their LLMs, the technology that underpins generative AI services such as ChatGPT.
Online search giant Baidu, TikTok owner ByteDance and social media and video gaming giant Tencent Holdings all slashed the prices of their LLM products, creating headwinds for China’s host of AI start-ups that are also trying to generate profits.
Moonshot, backed by investors including Alibaba, is among a group of prominent Chinese AI start-ups that have emerged as likely challengers to tech giants and OpenAI in the US.
Beijing Zhipu Huazhang Technology, established in 2019 to build on research from the Knowledge Engineering Group of Beijing’s prestigious Tsinghua University, is another closely watched domestic AI unicorn, referring to privately held start-ups valued at more than US$1 billion.
In June, Zhipu AI also said that it would cut the price of its GLM series of LLMs by 90 per cent, which it said was “not just a simple price war”, but represents its innovation.