iFlytek Spark Cognitive Large Model V3.0 officially released and benchmarked against ChatGPT 3.5
Today, the 2023 iFlytek Global 1024 Developer Festival opened in Hefei. iFlytek Chairman Liu Qingfeng announced the official release of iFlytek Spark 3.0, claiming that its comprehensive capabilities have surpassed ChatGPT (non-GPT-4 version).
According to reports, Spark Cognitive Large Model V3.0 has excellent performance in text generation (+7%), language understanding (+6%), knowledge question and answer (+7%), logical reasoning (8%), mathematical ability (3%), The seven major abilities of coding ability (+9%) and multi-modal ability (+7%) have continued to improve.
According to officials, the iFlytek Spark 3.0 general model will be fully benchmarked against ChatGPT. Its Chinese ability has surpassed ChatGPT and its English ability is equivalent to it. Liu Qingfeng revealed that iFlytek Spark 3.0 is far ahead of ChatGPT in terms of mathematical capabilities. In the fields of design solution generation, code completion, testing and error detection, Spark 3.0's capabilities have surpassed ChatGPT.
The author noticed that a horizontal evaluation report by the National Research Institute of Economics of the Development Research Center of the State Council, which was displayed at Liu Qingfeng’s site, showed that the current comprehensive capabilities of the Spark large model have surpassed ChatGPT, leading the country, and its performance is particularly outstanding in the education, medical, and legal industries. . "
Liu Qingfeng pointed out that large cognitive models are opening a new era of growth, and large models are increasingly relevant to everyone. Liu Qingfeng pointed out: "Under the current huge challenges and uncertainties faced by global politics, economy and military, general artificial intelligence is the most certain source of growth power in this era."
Large models are increasingly relevant to everyone. According to Liu Qingfeng, with the help of iFlytek's iFlyCode tool, programmers can complete the development work that originally took three months for platform migration in one month.
iFlytek said that there is still a gap between the current Spark cognitive large model and OpenAI's GPT-4 model, and a model that benchmarks GPT-4 will be launched in the first half of next year.
In addition, iFlytek said that since its release on August 15, iFlyCode, an intelligent programming assistant based on the Spark Code large model, has achieved in-depth docking applications with 107 institutions including JD Cloud and iSoftStone. In terms of internal efficiency improvement, according to iFlytek's internal usage data, this upgraded iFlyCodeV2.0 has a wider efficiency improvement range, which can achieve 50% efficiency improvement in the design stage, 37% and 44% efficiency improvement in the coding stage and testing stage respectively. .