Half a year after the wind of GM's large model, the question of "practicality" suddenly weighed on the head of every manufacturer.
Compared with whether it has large model capabilities, whether related applications can be implemented and whether they have the ability to solve business needs has become the gold standard for detecting the color of large models in addition to parameter competition.
On August 15, this afternoon, iFLYTEK held a press conference with the theme of "Liberating Productivity and Unleashing Imagination", and officially released the iFLYTEK Spark Model V2.0.
(Source: iFLYTEK)
It is reported that "iFLYTEK Spark Cognitive Model V2.0" focuses on breaking through code capabilities and multi-modal interaction capabilities on the basis of continuously improving general capabilities, and launches more diversified application landing products for teachers, schools, enterprises, developers and so on. In addition, the conference will also bring further commercial landing results of the Spark Cognitive Big Model in education, office and other fields.
As usual, let's first take a look at "iFLYTEK Spark Cognitive Large Model V2.0".
According to the official introduction, in addition to the continuous improvement of existing capabilities, this Spark V2.0 brings two major capabilities: code capabilities and multimodal capabilities.
According to the on-site demonstration, the code writing ability of Xinghuo V2.0 based on Python and C++ has been highly approaching ChatGPT, with a gap of only 1% and 2%, and has surpassed ChatGPT in the dimensions of code correction interpretation and single-person testing, and other capabilities are also catching up quickly.
(Source: iFLYTEK)
As for multimodal capabilities, iFLYTEK has made important progress in image understanding, picture description, picture reasoning, text and image generation, etc., significantly improving the understanding ability of large models to understand scenarios and intelligent support capabilities for life and industrial fields.
Based on this ability, the multimedia creative production platform "iFLYTEK Smart Work 2.0", which can greatly improve the efficiency of graphic creativity, video dubbing and video production, was released simultaneously, and readers who are in the mood are welcome to experience it.
(Source: iFLYTEK)
In addition to improving capabilities, iFLYTEK has not stopped the pace of promoting the landing of related capabilities.
(Source: iFLYTEK)
The "iFLYTEK Spark All-in-One Machine" launched by iFLYTEK and Huawei was officially released at this press conference.
(Source: iFLYTEK)
According to reports, iFLYTEK Xinghuo all-in-one machine makes all enterprises "more convenient, more autonomous, safer and controllable privatization and deployment of large model capabilities", and carries out privatization deployment on the platform of completely domestic independent innovation to protect their own data security.
Liu Qingfeng, Chairman of iFLYTEK, revealed that iFLYTEK Spark All-in-One Machine will provide independent and controllable privatization models based on industry-specific data for customers in key industries such as finance, state-owned enterprises, and central enterprises, accelerate the digital and intelligent transformation and upgrading of key industries, and have reached pilot intentions with a number of leading customers in the industry.
(Source: iFLYTEK)
From the content displayed at this conference, it is not difficult to see that iFLYTEK has a clear understanding and planning for the development of the large model, and the major strategies in the roadmap released in February this year are steadily being achieved.
According to the official roadmap of iFLYTEK, the iFLYTEK Spark model will usher in another upgrade on October 24, and the general model will be benchmarked against ChatGPT to achieve Chinese surpass and English equivalence.
As for surpassing ChatGPT? Maybe it's just a matter of time.