Compas.com – Perusahaan Teknologi Cina, Alibaba, meluncurkan model bahasa besar (LLM) alias dengan model kecerdasan buatan baru (AI) bernama Queen 2.5 -max.
This AI model has been directly published on the Chinese New Year. Quain 2.5 Max is a new version of Quain 2.5, published since September 2024.
Compared to the previous model, AI Alibaba Quain 2.5-Max was made with more token, 20 more tokens. By painting, Quain 2.5 is designed to use 18 billion tokens. With more tokens, AI models will have a better performance.
Also read: Alibaba AI Quane 2.5, open source and hundreds of models there
According to Alibaba, the Quain 2.5-Max has a high performance of other AI models including Deepsk-V3 in GPT-4O.
“Quain 2.5 Max GPT -4O, Deepsik -V3 and Lalama -3.1-405B are best in almost all aspects,” Alibaba Cloud said in an advertisement at the company.
Also Read: Deepsik costs cheap china that shocks the United States
Lebih khusus lagi, model AI Alibaba baru lebih tinggi di berbagai platform referensi, termasuk Arena-Durn, Live Bench, Live Roadbanch, MMLU dan GPQA-Diamond. Meven 2.5 Max Mawen Scores has also recorded a new record at MMLU and LifeWork Road, so it has become a sign of the SOP IA model.
In addition to sophistication, the AI model is in the output calendar spotlight, as it coincides with the Chinese New Year. Because generally, Chinese work and meet with family during celebrations.
Ini menunjukkan bahwa perusahaan teknis, terutama model AI, telah membantu mengembangkan Dello, terutama Deepsk, yang diskusi telah dibahas baru -baru ini, mencoba bersaing dengan produk yang sama dari para pesaingnya.
Deepsik is an artificial intelligence company in China that develops AI models to AI assistants like Open IA.
Also Read: What is Deepsk, China Chatgupt Challenger?
Deepske recently shocked people from its sophisticated AI models, such as other famous AI models, even with low cost.
Popularitas Deepsik baru -baru ini mengguncang pasar saham di Amerika Serikat, sehingga kapitalisasi pasar dari sejumlah perusahaan teknis, termasuk NVIDIA, telah dikalahkan.
As a result, Deepcin is considered a pressure on technical companies in the United States and China, reported by Reuters.
Also Read: AI comes to Dippeak, Nvidia Stocks Free
The latest model of AI published by Deepsik is called Deepsik R1 on January 20.
Two days later, with the release of Deepsk-R1, Mother Ticktock Bedentens began an update for the model of its headlights, called a double-5-Pro.
The update is supposed to make the AI model more than the last AI OpenAI model in terms of AIM. This measurement platform measures how the AI model understands and reacts to complex restrictions.
