Skip to content
I obtained an invitation code for Tongyi Qianwen from Alibaba Damo Academy during its public testing.
Below are the responses from Tongyi Qianwen, Wenxin Yiyan, ChatGPT (GPT 3.5), and ChatGPT (GPT 4) to the same question.
This comparison focuses on a single question and does not involve context. It allows for a direct comparison of the domestic giants against GPT in terms of strengths and weaknesses.
Two simple and intuitive personal judgments:
– Based on the performance of the two large models, since ChatGPT has already explored a path, the cost for later entrants to imitate is not as high as imagined, at least for large companies. With 1-2 months of tuning, they can reach a level that seems acceptable. After six months to a year, they may close the gap;
– Large companies with NLP accumulation move faster. The latter two, as long as they have money, also have opportunities. The arms race is beneficial for entrepreneurs and users, as costs will decrease, and finding valuable scenarios will be more worthwhile.
