On May 11th, Duozhi.com reported that iFlytek held two consecutive investor events on the 9th and 11th after the release of the Starfire model. During these events, they revealed the commercial landing plans for the Starfire model, corresponding timelines, and addressed the issue of the alleged ‘wrapper ChatGPT’ from OpenAI.
iFlytek mentioned that since the release of the Starfire model, more than 3,000 developers from 36 industries have expressed clear intentions for collaboration.
Regarding online rumors that the iFlytek Starfire model is a ‘wrapper ChatGPT’ from OpenAI, iFlytek responded that it is neither factual nor logical. They stated that if it were a wrapper ChatGPT, it would not be possible for the response speed of the iFlytek Starfire model to be faster than ChatGPT. Moreover, the results of the iFlytek Starfire model in text generation, knowledge inquiry, mathematical ability, and other aspects are superior to those of ChatGPT.
iFlytek revealed the specific iteration timeline for the Starfire model as follows:
June 9th: Breakthrough in open-ended question answering (real-time Q&A) by obtaining knowledge through a search plugin, presenting it in a more user-friendly manner. In addition, there will be an upgrade in multi-turn dialogue capability, and mathematical ability will reach a new level.
August 15th: Upgrade in code generation capability to enable developers and partners to use it efficiently and conveniently. iFlytek has already been using code generation internally. At the same time, multimodal interaction capability will be officially available to customers.
October 24th (iFlytek Global Developer Festival): Through scientific and systematic evaluation, the general cognitive model’s capabilities will be on par with ChatGPT in terms of surpassing it in Chinese and reaching a comparable level in English.”