iFlytek released the Spark model V4.0, officially challenging GPT-4 Turbo, and a competition about artificial intelligence in the field of education began. With the theme of "AI Assistant that Understands You", the conference focused on demonstrating the improvements of the Spark model in text generation, language understanding, logical reasoning, etc., and emphasized its advantages in practical applications, such as document traceability, Multi-language support and convenient interface interaction design. This release is not only a demonstration of technical strength, but also a bold innovation of iFlytek's traditional education model, which indicates that the application of artificial intelligence in the field of education will enter a new stage.
When the battle for AI ignites in the field of education, iFlytek comes out with the Spark model V4.0, and a battle about intelligence and wisdom officially begins. This is not only a leap in technology, but also a subversion of the traditional education model. At this conference with the theme of "AI Assistant that Understands You", iFlytek issued a challenge to GPT-4Turbo, vowing to compete on the AI battlefield.
Liu Qingfeng, the head of iFlytek, proudly announced that the Spark model has surpassed GPT-4 Turbo in terms of text generation, language understanding, logical reasoning and other capabilities. This is not only a breakthrough in technology, but also a tribute to human wisdom. The iFlytek Spark App has been downloaded more than 131 million times, smart hardware sales increased by 70% year-on-year, and the average monthly usage exceeds 40 million. This series of figures is enough to prove the strong strength of iFlytek in the AI field.
But numbers are just appearances, and the real battle unfolds in practical applications. The document traceability function of the iFlytek Spark model can scan documents and give answers, display reference sources, and support multi-lingual documents. This is not only respect for knowledge, but also a deep understanding of learners’ needs. The one-sentence reproduction, speech recognition in high-noise scenes, and multi-dialect and multi-language switching-free functions take iFlytek Spark to a new level of practicality.
In the actual test, the performance of the iFlytek Spark model was equally impressive. Whether it is in the logical paragraphs of the mentally retarded bar or in abstract literary creation, iFlytek Spark has shown considerable strength. Although there is still room for improvement in humor understanding and creative writing, this does not prevent it from winning the recognition of testers with rigorous output feedback in multi-modal ability tests.
Mathematics, as a shortcoming of large models, iFlytek Spark performs well on simple questions when dealing with college entrance examination mathematics questions, but it seems a little inadequate on more difficult questions. However, this does not affect its excellent performance in summarizing long documents and generating PPT with one click. The interface interaction design of iFlytek Spark is intuitive and easy to use, which greatly improves work efficiency.
In 2024, the competition for large models has shifted from the technical level to the application and product implementation. Liu Qingfeng, Chairman of iFlytek, emphasized the importance of the implementation of large models and believed that how to realize the rigidly needed value of large models is now the top priority for large model companies. The iFlytek Spark large model V4 has been logged into the iFlytek Spark App on Android and iOS, and the web version has also been updated simultaneously. For the business end, iFlytek develops customized large model solutions based on the characteristics and needs of different industries, enabling the model to be applied in multiple fields.
All in all, the release of iFlytek Spark Model V4.0 marks that AI competition in the education field has entered a fierce stage. Its innovation at the application level and emphasis on user needs are worthy of the industry's attention and learning. In the future, whether Xinghuo Large Model can continue to maintain its leading advantage remains to be tested by time.